{"data":[{"alias":"wan2.7-image","display_name":"wan2.7-image","description":"","category":"image","categories":["image"],"capabilities":["image"],"input_modalities":null,"supports_structured_output":false,"is_chinese_model":false,"channels":[{"input_per_million_tokens":0,"output_per_million_tokens":0,"per_call_price":0.03}],"discount_multiplier":1,"discount_source":"none"},{"alias":"dreamina-seedance-2-0-260128","display_name":"dreamina-seedance-2-0-260128","description":"","category":"video","categories":["video"],"capabilities":["video"],"input_modalities":null,"supports_structured_output":false,"is_chinese_model":false,"channels":[{"input_per_million_tokens":0,"output_per_million_tokens":7,"video_output_per_million_tokens":7,"video_output_per_million_tokens_by_resolution":{"1080p":7.7,"4k":4}}],"discount_multiplier":1,"discount_source":"none"},{"alias":"whisper-1","display_name":"Whisper v1","description":"OpenAI's speech recognition model. Supports timestamps. $0.006/min.","category":"audio","categories":["audio"],"capabilities":["transcription"],"input_modalities":["audio"],"supports_structured_output":false,"is_chinese_model":false,"channels":[{"input_per_million_tokens":0,"output_per_million_tokens":0,"per_call_price":0.006}],"discount_multiplier":1,"discount_source":"none"},{"alias":"qwen-image-3.0","display_name":"qwen-image-3.0","description":"","category":"image","categories":["image"],"capabilities":["image"],"input_modalities":null,"supports_structured_output":false,"is_chinese_model":false,"channels":[{"input_per_million_tokens":0,"output_per_million_tokens":0,"per_call_price":0.03,"per_input_image_price":0.003}],"discount_multiplier":1,"discount_source":"none"},{"alias":"chirp-2","display_name":"chirp-2","description":"Google Chirp 2 — dedicated multilingual ASR (USM). Byte upload, duration-billed.","category":"audio","categories":["audio"],"capabilities":["transcription"],"input_modalities":["audio"],"supports_structured_output":false,"is_chinese_model":false,"channels":[{"input_per_million_tokens":0,"output_per_million_tokens":0,"per_call_price":0.016}],"discount_multiplier":1,"discount_source":"none"},{"alias":"chirp-3","display_name":"chirp-3","description":"Google Chirp 3 — latest USM ASR. Enhanced multilingual accuracy, diarization, auto language. Byte upload, duration-billed.","category":"audio","categories":["audio"],"capabilities":["transcription"],"input_modalities":["audio"],"supports_structured_output":false,"is_chinese_model":false,"channels":[{"input_per_million_tokens":0,"output_per_million_tokens":0,"per_call_price":0.016}],"discount_multiplier":1,"discount_source":"none"},{"alias":"veo-3.1-generate-001","display_name":"veo-3.1-generate-001","description":"","category":"video","categories":["video"],"capabilities":["video"],"input_modalities":null,"supports_structured_output":false,"is_chinese_model":false,"channels":[{"input_per_million_tokens":0,"output_per_million_tokens":0,"video_per_second_usd":0.4,"video_per_second_no_audio_usd":0.2}],"discount_multiplier":1,"discount_source":"none"},{"alias":"paraformer-realtime-v2","display_name":"paraformer-realtime-v2","description":"Alibaba Paraformer v2 real-time streaming speech recognition. Cheapest tier; Chinese + many dialects + 7 langs.","category":"audio","categories":["audio"],"capabilities":["transcription"],"input_modalities":["audio"],"supports_structured_output":false,"is_chinese_model":false,"channels":[{"input_per_million_tokens":0,"output_per_million_tokens":0,"per_call_price":0.002}],"discount_multiplier":1,"discount_source":"none"},{"alias":"veo-3.1-fast-generate-001","display_name":"veo-3.1-fast-generate-001","description":"","category":"video","categories":["video"],"capabilities":["video"],"input_modalities":null,"supports_structured_output":false,"is_chinese_model":false,"channels":[{"input_per_million_tokens":0,"output_per_million_tokens":0,"video_per_second_usd":0.15,"video_per_second_no_audio_usd":0.1}],"discount_multiplier":1,"discount_source":"none"},{"alias":"veo-2.0-generate-001","display_name":"veo-2.0-generate-001","description":"","category":"video","categories":["video"],"capabilities":["video"],"input_modalities":null,"supports_structured_output":false,"is_chinese_model":false,"channels":[{"input_per_million_tokens":0,"output_per_million_tokens":0,"video_per_second_usd":0.5}],"discount_multiplier":1,"discount_source":"none"},{"alias":"sora-2-pro","display_name":"sora-2-pro","description":"","category":"video","categories":["video"],"capabilities":["video"],"input_modalities":null,"supports_structured_output":false,"is_chinese_model":false,"channels":[{"input_per_million_tokens":0,"output_per_million_tokens":0,"video_per_second_usd":0.3,"video_per_second_usd_by_resolution":{"1024p":0.5,"720p":0.3}}],"discount_multiplier":1,"discount_source":"none"},{"alias":"sora-2","display_name":"sora-2","description":"","category":"video","categories":["video"],"capabilities":["video"],"input_modalities":null,"supports_structured_output":false,"is_chinese_model":false,"channels":[{"input_per_million_tokens":0,"output_per_million_tokens":0,"video_per_second_usd":0.1}],"discount_multiplier":1,"discount_source":"none"},{"alias":"seedance-1-5-pro-251215","display_name":"seedance-1-5-pro-251215","description":"","category":"video","categories":["video"],"capabilities":["video"],"input_modalities":null,"supports_structured_output":false,"is_chinese_model":false,"channels":[{"input_per_million_tokens":0,"output_per_million_tokens":2.4,"video_output_per_million_tokens":2.4,"video_output_no_audio_per_million_tokens":1.2}],"discount_multiplier":1,"discount_source":"none"},{"alias":"seedance-1-0-pro-fast-251015","display_name":"seedance-1-0-pro-fast-251015","description":"","category":"video","categories":["video"],"capabilities":["video"],"input_modalities":null,"supports_structured_output":false,"is_chinese_model":false,"channels":[{"input_per_million_tokens":0,"output_per_million_tokens":1,"video_output_per_million_tokens":1}],"discount_multiplier":1,"discount_source":"none"},{"alias":"qwen-image-2.0-pro","display_name":"qwen-image-2.0-pro","description":"","category":"image","categories":["image"],"capabilities":["image"],"input_modalities":null,"supports_structured_output":false,"is_chinese_model":false,"channels":[{"input_per_million_tokens":0,"output_per_million_tokens":0,"per_call_price":0.075}],"discount_multiplier":1,"discount_source":"none"},{"alias":"dreamina-seedance-2-0-mini-260615","display_name":"dreamina-seedance-2-0-mini-260615","description":"","category":"video","categories":["video"],"capabilities":["video"],"input_modalities":null,"supports_structured_output":false,"is_chinese_model":false,"channels":[{"input_per_million_tokens":0,"output_per_million_tokens":3.5,"video_output_per_million_tokens":3.5}],"discount_multiplier":1,"discount_source":"none"},{"alias":"seed-asr-bigmodel","display_name":"seed-asr-bigmodel","description":"ByteDance Seed offline audio-file ASR. Async submit/poll, URL input, duration-billed. Speaker diarization supported: response_format=diarized_json → per-segment speaker labels (enable_speaker_info).","category":"audio","categories":["audio"],"capabilities":["transcription"],"input_modalities":["audio"],"supports_structured_output":false,"is_chinese_model":false,"channels":[{"input_per_million_tokens":0,"output_per_million_tokens":0,"per_call_price":0.002}],"discount_multiplier":1,"discount_source":"none"},{"alias":"seedream-4-0-250828","display_name":"seedream-4-0-250828","description":"","category":"image","categories":["image"],"capabilities":["image"],"input_modalities":null,"supports_structured_output":false,"is_chinese_model":false,"channels":[{"input_per_million_tokens":0,"output_per_million_tokens":0,"per_call_price":0.03}],"discount_multiplier":1,"discount_source":"none"},{"alias":"dreamina-seedance-2-0-fast-260128","display_name":"dreamina-seedance-2-0-fast-260128","description":"","category":"video","categories":["video"],"capabilities":["video"],"input_modalities":null,"supports_structured_output":false,"is_chinese_model":false,"channels":[{"input_per_million_tokens":0,"output_per_million_tokens":5.6,"video_output_per_million_tokens":5.6}],"discount_multiplier":1,"discount_source":"none"},{"alias":"seedream-4-5-251128","display_name":"seedream-4-5-251128","description":"","category":"image","categories":["image"],"capabilities":["image"],"input_modalities":null,"supports_structured_output":false,"is_chinese_model":false,"channels":[{"input_per_million_tokens":0,"output_per_million_tokens":0,"per_call_price":0.04}],"discount_multiplier":1,"discount_source":"none"},{"alias":"seedream-5-0-260128","display_name":"seedream-5-0-260128","description":"","category":"image","categories":["image"],"capabilities":["image"],"input_modalities":null,"supports_structured_output":false,"is_chinese_model":false,"channels":[{"input_per_million_tokens":0,"output_per_million_tokens":0,"per_call_price":0.035}],"discount_multiplier":1,"discount_source":"none"},{"alias":"qwen3-asr-flash","display_name":"Qwen3-ASR Flash","description":"Qwen3-ASR Flash (Alibaba). OpenAI-compatible chat/input_audio; duration-billed via upstream usage.seconds (exact).","category":"audio","categories":["audio"],"capabilities":["transcription"],"input_modalities":["audio"],"supports_structured_output":false,"is_chinese_model":true,"channels":[{"input_per_million_tokens":0,"output_per_million_tokens":0,"per_call_price":0.0021}],"discount_multiplier":1,"discount_source":"none"},{"alias":"fun-asr","display_name":"Fun-ASR (recording file)","description":"Alibaba Fun-ASR recording-file recognition. Async offline ASR; supports speaker diarization (response_format=diarized_json → per-sentence speaker). URL or file input; duration-billed via upstream usage.duration (exact). Chinese-first (mandarin + cantonese + dialects).","category":"audio","categories":["audio"],"capabilities":["transcription"],"input_modalities":["audio"],"supports_structured_output":false,"is_chinese_model":true,"channels":[{"input_per_million_tokens":0,"output_per_million_tokens":0,"per_call_price":0.0021}],"discount_multiplier":1,"discount_source":"none"},{"alias":"fun-asr-flash","display_name":"Fun-ASR Flash","description":"Alibaba Fun-ASR Flash. Base64 input_audio, sync, duration-billed via upstream usage.duration (exact). Chinese-first.","category":"audio","categories":["audio"],"capabilities":["transcription"],"input_modalities":["audio"],"supports_structured_output":false,"is_chinese_model":true,"channels":[{"input_per_million_tokens":0,"output_per_million_tokens":0,"per_call_price":0.0021}],"discount_multiplier":1,"discount_source":"none"},{"alias":"fun-asr-mtl","display_name":"Fun-ASR MTL (multilingual, recording file)","description":"Alibaba Fun-ASR MTL multilingual recording-file recognition. Async offline ASR; supports speaker diarization (response_format=diarized_json → per-sentence speaker). URL or file input; duration-billed via upstream usage.duration (exact).","category":"audio","categories":["audio"],"capabilities":["transcription"],"input_modalities":["audio"],"supports_structured_output":false,"is_chinese_model":false,"channels":[{"input_per_million_tokens":0,"output_per_million_tokens":0,"per_call_price":0.0021}],"discount_multiplier":1,"discount_source":"none"},{"alias":"fun-asr-realtime","display_name":"fun-asr-realtime","description":"Alibaba Fun-ASR real-time streaming speech recognition. WebSocket, wall-clock billed. Chinese-first with hotwords; server-side VAD.","category":"audio","categories":["audio"],"capabilities":["transcription"],"input_modalities":["audio"],"supports_structured_output":false,"is_chinese_model":false,"channels":[{"input_per_million_tokens":0,"output_per_million_tokens":0,"per_call_price":0.002}],"discount_multiplier":1,"discount_source":"none"},{"alias":"wan2.7-image-pro","display_name":"wan2.7-image-pro","description":"","category":"image","categories":["image"],"capabilities":["image"],"input_modalities":null,"supports_structured_output":false,"is_chinese_model":false,"channels":[{"input_per_million_tokens":0,"output_per_million_tokens":0,"per_call_price":0.075}],"discount_multiplier":1,"discount_source":"none"},{"alias":"qwen-image-2.0","display_name":"qwen-image-2.0","description":"","category":"image","categories":["image"],"capabilities":["image"],"input_modalities":null,"supports_structured_output":false,"is_chinese_model":false,"channels":[{"input_per_million_tokens":0,"output_per_million_tokens":0,"per_call_price":0.035}],"discount_multiplier":1,"discount_source":"none"},{"alias":"qwen3-asr-flash-realtime","display_name":"qwen3-asr-flash-realtime","description":"Alibaba Qwen3-ASR-Flash real-time streaming speech recognition. Strongest Chinese (mandarin + cantonese + dialects + emotion) + 30+ languages.","category":"audio","categories":["audio"],"capabilities":["transcription"],"input_modalities":["audio"],"supports_structured_output":false,"is_chinese_model":false,"channels":[{"input_per_million_tokens":0,"output_per_million_tokens":0,"per_call_price":0.002}],"discount_multiplier":1,"discount_source":"none"},{"alias":"qwen3-vl-flash","display_name":"Qwen3 VL Flash","description":"Fast, low-cost Qwen vision + thinking multimodal model.","category":"chat","categories":["chat"],"max_input_tokens":256000,"max_output_tokens":16384,"capabilities":["chat","vision","reasoning","tools"],"input_modalities":["text","image","video"],"supports_structured_output":true,"is_chinese_model":true,"vs_gpt4o_saving":"99%+","channels":[{"input_per_million_tokens":0.05,"output_per_million_tokens":0.4,"cache_read_input_per_million_tokens":0.022,"tiers":[{"max_input_tokens":32000,"input_per_m":0.05,"output_per_m":0.4},{"max_input_tokens":128000,"input_per_m":0.075,"output_per_m":0.6},{"max_input_tokens":256000,"input_per_m":0.12,"output_per_m":0.96}]}],"discount_multiplier":1,"discount_source":"none"},{"alias":"nova-2-sonic","display_name":"nova-2-sonic","description":"Amazon Nova 2 Sonic speech-to-speech over Bedrock bidirectional streaming (/v1/realtime). Native S2S, top speech-reasoning benchmark. Audio $3/M in · $12/M out; text $0.06/M in · $0.24/M out. Billed per response at official list price.","category":"realtime","categories":["realtime"],"capabilities":["realtime","audio"],"input_modalities":["audio","text"],"supports_structured_output":false,"is_chinese_model":false,"vs_gpt4o_saving":"99%","channels":[{"input_per_million_tokens":0.06,"output_per_million_tokens":0.24}],"release_stage":"invited_beta","discount_multiplier":1,"discount_source":"none"},{"alias":"gemini-2.5-flash-lite","display_name":"Gemini 2.5 Flash-Lite","description":"Ultra-low cost and latency model for high-volume workloads.","category":"chat","categories":["chat"],"max_input_tokens":1048576,"max_output_tokens":65536,"capabilities":["chat","vision","code","tools"],"input_modalities":["text","image","audio","video","file"],"supports_structured_output":true,"is_chinese_model":false,"vs_gpt4o_saving":"98%","channels":[{"input_per_million_tokens":0.1,"output_per_million_tokens":0.4,"cache_read_input_per_million_tokens":0.01},{"input_per_million_tokens":0.1,"output_per_million_tokens":0.4,"cache_read_input_per_million_tokens":0.01}],"discount_multiplier":1,"discount_source":"none"},{"alias":"Dola-Seed-2.0-mini","display_name":"Dola Seed 2.0 Mini","description":"Smallest / cheapest Seed 2.0 variant for classification / extraction.","category":"chat","categories":["chat"],"max_input_tokens":262144,"max_output_tokens":4096,"capabilities":["chat","tools"],"input_modalities":["text"],"supports_structured_output":true,"is_chinese_model":true,"vs_gpt4o_saving":"98%","channels":[{"input_per_million_tokens":0.1,"output_per_million_tokens":0.4,"cache_read_input_per_million_tokens":0.02,"tiers":[{"max_input_tokens":128000,"input_per_m":0.1,"output_per_m":0.4},{"input_per_m":0.2,"output_per_m":0.8}]}],"discount_multiplier":1,"discount_source":"none"},{"alias":"qwen3.5-flash","display_name":"Qwen3.5 Flash","description":"High-throughput, low-cost Qwen text model.","category":"chat","categories":["chat"],"max_input_tokens":1048576,"max_output_tokens":16384,"capabilities":["chat","code","reasoning","tools"],"input_modalities":["text"],"supports_structured_output":true,"is_chinese_model":true,"vs_gpt4o_saving":"98%","channels":[{"input_per_million_tokens":0.1,"output_per_million_tokens":0.4,"cache_read_input_per_million_tokens":0.029}],"discount_multiplier":1,"discount_source":"none"},{"alias":"qwen3-coder-next","display_name":"Qwen3 Coder Next","description":"Alibaba's next-gen Qwen3 coding model with enhanced reasoning and tool use.","category":"chat","categories":["chat"],"max_input_tokens":262144,"max_output_tokens":64000,"capabilities":["chat","code","reasoning","tools","long-context"],"input_modalities":["text"],"supports_structured_output":true,"is_chinese_model":true,"vs_gpt4o_saving":"98%","channels":[{"input_per_million_tokens":0.11,"output_per_million_tokens":0.8,"cache_read_input_per_million_tokens":0.022}],"release_stage":"invited_beta","discount_multiplier":1,"discount_source":"none"},{"alias":"deepseek-v4-flash","display_name":"DeepSeek V4 Flash","description":"DeepSeek's fast, low-cost V4 Flash model.","category":"chat","categories":["chat"],"max_input_tokens":1000000,"max_output_tokens":8192,"capabilities":["chat","code","tools"],"input_modalities":["text"],"supports_structured_output":true,"is_chinese_model":true,"vs_gpt4o_saving":"97%","channels":[{"input_per_million_tokens":0.138,"output_per_million_tokens":0.275,"cache_read_input_per_million_tokens":0.0028},{"input_per_million_tokens":0.138,"output_per_million_tokens":0.275,"cache_read_input_per_million_tokens":0.0028},{"input_per_million_tokens":0.138,"output_per_million_tokens":0.275,"cache_read_input_per_million_tokens":0.0028}],"discount_multiplier":1,"discount_source":"none"},{"alias":"hunyuan-3","display_name":"Hunyuan 3","description":"Tencent's Hunyuan 3 large language model; 256K context, long-form Chinese and English reasoning.","category":"chat","categories":["chat"],"max_input_tokens":262144,"max_output_tokens":128000,"capabilities":["chat","code","tools","reasoning"],"input_modalities":["text"],"supports_structured_output":true,"is_chinese_model":true,"vs_gpt4o_saving":"97%","channels":[{"input_per_million_tokens":0.14,"output_per_million_tokens":0.58,"cache_read_input_per_million_tokens":0.035}],"release_stage":"invited_beta","discount_multiplier":1,"discount_source":"none"},{"alias":"glm-5.3-flash","display_name":"GLM-5.3-Flash","description":"Z.ai's natively multimodal GLM-5.3-Flash: image and video input, always-on thinking, 1M-token context.","category":"chat","categories":["chat"],"max_input_tokens":1048576,"max_output_tokens":163840,"capabilities":["chat","code","reasoning","vision","tools","long-context"],"input_modalities":["text","image","video"],"supports_structured_output":true,"is_chinese_model":true,"vs_gpt4o_saving":"97%","channels":[{"input_per_million_tokens":0.15,"output_per_million_tokens":0.5,"cache_read_input_per_million_tokens":0.03}],"discount_multiplier":1,"discount_source":"none"},{"alias":"qwen3-vl-plus","display_name":"Qwen3 VL Plus","description":"Qwen vision + thinking multimodal model.","category":"chat","categories":["chat"],"max_input_tokens":256000,"max_output_tokens":16384,"capabilities":["chat","vision","reasoning","tools"],"input_modalities":["text","image","video"],"supports_structured_output":true,"is_chinese_model":true,"vs_gpt4o_saving":"96%","channels":[{"input_per_million_tokens":0.2,"output_per_million_tokens":1.6,"cache_read_input_per_million_tokens":0.143,"tiers":[{"max_input_tokens":32000,"input_per_m":0.2,"output_per_m":1.6},{"max_input_tokens":128000,"input_per_m":0.3,"output_per_m":2.4},{"max_input_tokens":256000,"input_per_m":0.6,"output_per_m":4.8}]}],"discount_multiplier":1,"discount_source":"none"},{"alias":"gpt-5.4-nano","display_name":"GPT-5.4 Nano","description":"OpenAI's fastest and most affordable model for classification, extraction, and simple tasks.","category":"chat","categories":["chat"],"max_input_tokens":400000,"max_output_tokens":128000,"capabilities":["chat","code","tools","reasoning"],"input_modalities":["text","image","file"],"supports_structured_output":true,"is_chinese_model":false,"vs_gpt4o_saving":"96%","channels":[{"input_per_million_tokens":0.2,"output_per_million_tokens":1.25,"cache_read_input_per_million_tokens":0.1}],"discount_multiplier":1,"discount_source":"none"},{"alias":"minimax-m2.5","display_name":"MiniMax M2.5","description":"MiniMax M2.5 agentic/coding model. ~205K context.","category":"chat","categories":["chat"],"max_input_tokens":204800,"max_output_tokens":64000,"capabilities":["chat","code","reasoning","tools"],"input_modalities":["text"],"supports_structured_output":true,"is_chinese_model":true,"vs_gpt4o_saving":"96%","channels":[{"input_per_million_tokens":0.22,"output_per_million_tokens":0.9,"cache_read_input_per_million_tokens":0.044}],"release_stage":"invited_beta","discount_multiplier":1,"discount_source":"none"},{"alias":"gemini-3.1-flash-lite-preview","display_name":"Gemini 3.1 Flash-Lite","description":"Google's most cost-effective model with 1M context window.","category":"chat","categories":["chat"],"max_input_tokens":1048576,"max_output_tokens":65536,"capabilities":["chat","vision","code","tools"],"input_modalities":["text","image","audio","video","file"],"supports_structured_output":true,"is_chinese_model":false,"vs_gpt4o_saving":"95%","channels":[{"input_per_million_tokens":0.25,"output_per_million_tokens":1.5},{"input_per_million_tokens":0.25,"output_per_million_tokens":1.5}],"discount_multiplier":1,"discount_source":"none"},{"alias":"ByteDance-Seed-1.8","display_name":"ByteDance Seed 1.8","description":"Previous-generation Seed model retained for backwards compatibility.","category":"chat","categories":["chat"],"max_input_tokens":262144,"max_output_tokens":4096,"capabilities":["chat","tools"],"input_modalities":["text"],"supports_structured_output":true,"is_chinese_model":true,"vs_gpt4o_saving":"95%","channels":[{"input_per_million_tokens":0.25,"output_per_million_tokens":2,"cache_read_input_per_million_tokens":0.05,"tiers":[{"max_input_tokens":128000,"input_per_m":0.25,"output_per_m":2},{"input_per_m":0.5,"output_per_m":4}]}],"discount_multiplier":1,"discount_source":"none"},{"alias":"qwen3.6-flash","display_name":"Qwen3.6 Flash","description":"Fast, low-cost Qwen3.6 model.","category":"chat","categories":["chat"],"max_input_tokens":256000,"max_output_tokens":32768,"capabilities":["chat","code","tools"],"input_modalities":["text"],"supports_structured_output":true,"is_chinese_model":true,"vs_gpt4o_saving":"95%","channels":[{"input_per_million_tokens":0.25,"output_per_million_tokens":1.5,"cache_read_input_per_million_tokens":0.05}],"discount_multiplier":1,"discount_source":"none"},{"alias":"gemini-3.1-flash-lite-image","display_name":"gemini-3.1-flash-lite-image","description":"","category":"image","categories":["image"],"capabilities":["image"],"input_modalities":null,"supports_structured_output":false,"is_chinese_model":false,"vs_gpt4o_saving":"95%","channels":[{"input_per_million_tokens":0.25,"output_per_million_tokens":30}],"discount_multiplier":1,"discount_source":"none"},{"alias":"Dola-Seed-2.0-lite","display_name":"Dola Seed 2.0 Lite","description":"Lower-latency tier of Seed 2.0 for high-volume workloads.","category":"chat","categories":["chat"],"max_input_tokens":262144,"max_output_tokens":4096,"capabilities":["chat","code","tools"],"input_modalities":["text"],"supports_structured_output":true,"is_chinese_model":true,"vs_gpt4o_saving":"95%","channels":[{"input_per_million_tokens":0.25,"output_per_million_tokens":2,"cache_read_input_per_million_tokens":0.05,"tiers":[{"max_input_tokens":128000,"input_per_m":0.25,"output_per_m":2},{"input_per_m":0.5,"output_per_m":4}]}],"discount_multiplier":1,"discount_source":"none"},{"alias":"deepseek-3.2","display_name":"DeepSeek V3.2","description":"DeepSeek V3.2 MoE model with sparse attention; strong code and reasoning at low cost.","category":"chat","categories":["chat"],"max_input_tokens":128000,"max_output_tokens":64000,"capabilities":["chat","code","reasoning","tools"],"input_modalities":["text"],"supports_structured_output":true,"is_chinese_model":true,"vs_gpt4o_saving":"94%","channels":[{"input_per_million_tokens":0.28,"output_per_million_tokens":0.42,"cache_read_input_per_million_tokens":0.028}],"release_stage":"invited_beta","discount_multiplier":1,"discount_source":"none"},{"alias":"minimax-m2.1","display_name":"MiniMax M2.1","description":"MiniMax M2.1 agentic/coding model.","category":"chat","categories":["chat"],"max_input_tokens":204800,"max_output_tokens":64000,"capabilities":["chat","code","reasoning","tools"],"input_modalities":["text"],"supports_structured_output":true,"is_chinese_model":true,"vs_gpt4o_saving":"94%","channels":[{"input_per_million_tokens":0.3,"output_per_million_tokens":1.2,"cache_read_input_per_million_tokens":0.06}],"release_stage":"invited_beta","discount_multiplier":1,"discount_source":"none"},{"alias":"gemini-2.5-flash","display_name":"Gemini 2.5 Flash","description":"Fast, versatile model with thinking support and excellent cost-performance.","category":"chat","categories":["chat"],"max_input_tokens":1048576,"max_output_tokens":65536,"capabilities":["chat","vision","code","tools","reasoning"],"input_modalities":["text","image","audio","video","file"],"supports_structured_output":true,"is_chinese_model":false,"vs_gpt4o_saving":"94%","channels":[{"input_per_million_tokens":0.3,"output_per_million_tokens":2.5,"cache_read_input_per_million_tokens":0.03},{"input_per_million_tokens":0.3,"output_per_million_tokens":2.5,"cache_read_input_per_million_tokens":0.03}],"discount_multiplier":1,"discount_source":"none"},{"alias":"gemini-2.5-flash-image","display_name":"gemini-2.5-flash-image","description":"","category":"image","categories":["image"],"capabilities":["image"],"input_modalities":null,"supports_structured_output":false,"is_chinese_model":false,"vs_gpt4o_saving":"94%","channels":[{"input_per_million_tokens":0.3,"output_per_million_tokens":30}],"discount_multiplier":1,"discount_source":"none"},{"alias":"gemini-3.5-flash-lite","display_name":"Gemini 3.5 Flash-Lite","description":"Ultra-low-cost, low-latency Gemini 3.5 Flash-Lite with 1M context.","category":"chat","categories":["chat"],"max_input_tokens":1048576,"max_output_tokens":65536,"capabilities":["chat","vision","code","tools"],"input_modalities":["text","image","audio","video","file"],"supports_structured_output":true,"is_chinese_model":false,"vs_gpt4o_saving":"94%","channels":[{"input_per_million_tokens":0.3,"output_per_million_tokens":2.5,"cache_read_input_per_million_tokens":0.03}],"discount_multiplier":1,"discount_source":"none"},{"alias":"qwen3.5-plus","display_name":"Qwen3.5 Plus","description":"Cost-effective Qwen text model.","category":"chat","categories":["chat"],"max_input_tokens":1048576,"max_output_tokens":16384,"capabilities":["chat","code","reasoning","tools"],"input_modalities":["text"],"supports_structured_output":true,"is_chinese_model":true,"vs_gpt4o_saving":"92%","channels":[{"input_per_million_tokens":0.4,"output_per_million_tokens":2.4,"cache_read_input_per_million_tokens":0.115,"tiers":[{"max_input_tokens":256000,"input_per_m":0.4,"output_per_m":2.4},{"max_input_tokens":1048576,"input_per_m":0.5,"output_per_m":3}]}],"discount_multiplier":1,"discount_source":"none"},{"alias":"qwen3.7-plus","display_name":"Qwen3.7 Plus","description":"Balanced Qwen3.7 model for everyday tasks. 1M-token context.","category":"chat","categories":["chat"],"max_input_tokens":1000000,"max_output_tokens":65536,"capabilities":["chat","code","reasoning","tools","long-context"],"input_modalities":["text"],"supports_structured_output":true,"is_chinese_model":true,"vs_gpt4o_saving":"92%","channels":[{"input_per_million_tokens":0.4,"output_per_million_tokens":1.6,"cache_read_input_per_million_tokens":0.08}],"discount_multiplier":1,"discount_source":"none"},{"alias":"deepseek-v4-flash-0731","display_name":"DeepSeek V4 Flash (0731)","description":"DeepSeek V4 Flash 2026-07-31 snapshot. 1M context, reasoning-heavy; priced above the rolling deepseek-v4-flash.","category":"chat","categories":["chat"],"max_input_tokens":1000000,"max_output_tokens":393216,"capabilities":["chat","code","reasoning","tools"],"input_modalities":["text"],"supports_structured_output":true,"is_chinese_model":true,"vs_gpt4o_saving":"91%","channels":[{"input_per_million_tokens":0.44,"output_per_million_tokens":1.32,"cache_read_input_per_million_tokens":0.044},{"input_per_million_tokens":0.44,"output_per_million_tokens":1.32,"cache_read_input_per_million_tokens":0.044}],"discount_multiplier":1,"discount_source":"none"},{"alias":"gemini-3.1-flash-image","display_name":"gemini-3.1-flash-image","description":"","category":"image","categories":["image"],"capabilities":["image"],"input_modalities":null,"supports_structured_output":false,"is_chinese_model":false,"vs_gpt4o_saving":"90%","channels":[{"input_per_million_tokens":0.5,"output_per_million_tokens":60}],"discount_multiplier":1,"discount_source":"none"},{"alias":"Dola-Seed-2.0-pro","display_name":"Dola Seed 2.0 Pro","description":"ByteDance Seed 2.0 flagship — reasoning + tool use.","category":"chat","categories":["chat"],"max_input_tokens":262144,"max_output_tokens":4096,"capabilities":["chat","code","reasoning","tools"],"input_modalities":["text"],"supports_structured_output":true,"is_chinese_model":true,"vs_gpt4o_saving":"90%","channels":[{"input_per_million_tokens":0.5,"output_per_million_tokens":3,"cache_read_input_per_million_tokens":0.1,"tiers":[{"max_input_tokens":128000,"input_per_m":0.5,"output_per_m":3},{"input_per_m":1,"output_per_m":6}]}],"discount_multiplier":1,"discount_source":"none"},{"alias":"gemini-2.5-flash-native-audio","display_name":"gemini-2.5-flash-native-audio","description":"Google Gemini 2.5 Flash Native Audio speech-to-speech over WebSocket (AI Studio). Most natural native-audio voice. Audio $3/M in · $12/M out; text $0.50/M in · $2/M out. Billed per response at official list price.","category":"realtime","categories":["realtime"],"capabilities":["realtime","audio"],"input_modalities":["audio","text"],"supports_structured_output":false,"is_chinese_model":false,"vs_gpt4o_saving":"90%","channels":[{"input_per_million_tokens":0.5,"output_per_million_tokens":2}],"release_stage":"invited_beta","discount_multiplier":1,"discount_source":"none"},{"alias":"gemini-3-flash-preview","display_name":"Gemini 3 Flash","description":"Fast and capable model with configurable reasoning levels.","category":"chat","categories":["chat"],"max_input_tokens":1048576,"max_output_tokens":65536,"capabilities":["chat","vision","code","tools","reasoning","image"],"input_modalities":["text","image","audio","video","file"],"supports_structured_output":true,"is_chinese_model":false,"vs_gpt4o_saving":"90%","channels":[{"input_per_million_tokens":0.5,"output_per_million_tokens":3,"cache_read_input_per_million_tokens":0.05},{"input_per_million_tokens":0.5,"output_per_million_tokens":3,"cache_read_input_per_million_tokens":0.05}],"discount_multiplier":1,"discount_source":"none"},{"alias":"kimi-k2.5","display_name":"Kimi K2.5","description":"Moonshot's flagship multimodal model. Vision + text, native multimodal architecture.","category":"chat","categories":["chat"],"max_input_tokens":262144,"max_output_tokens":32768,"capabilities":["chat","vision","code","reasoning","tools"],"input_modalities":["text","image"],"supports_structured_output":true,"is_chinese_model":true,"vs_gpt4o_saving":"89%","channels":[{"input_per_million_tokens":0.574,"output_per_million_tokens":3.011,"cache_read_input_per_million_tokens":0.115}],"discount_multiplier":1,"discount_source":"none"},{"alias":"gpt-realtime-2.1-mini","display_name":"GPT Realtime 2.1 Mini","description":"Cost-efficient speech-to-speech realtime model over WebSocket (/v1/realtime). Audio $10/M in · $20/M out; text $0.60/M in · $2.40/M out; cached text $0.06/M, cached audio $0.30/M. Billed per response at official list price.","category":"realtime","categories":["realtime"],"capabilities":["realtime","audio"],"input_modalities":["audio","text"],"supports_structured_output":false,"is_chinese_model":false,"vs_gpt4o_saving":"88%","channels":[{"input_per_million_tokens":0.6,"output_per_million_tokens":2.4,"cache_read_input_per_million_tokens":0.06},{"input_per_million_tokens":0.6,"output_per_million_tokens":2.4,"cache_read_input_per_million_tokens":0.06}],"release_stage":"invited_beta","discount_multiplier":1,"discount_source":"none"},{"alias":"gemini-3.7-flash","display_name":"Gemini 3.7 Flash","description":"Google's most capable Flash model for agentic workflows and multimodal reasoning.","category":"chat","categories":["chat"],"max_input_tokens":1048576,"max_output_tokens":65536,"capabilities":["chat","vision","code","tools","reasoning"],"input_modalities":["text","image","audio","video","file"],"supports_structured_output":true,"is_chinese_model":false,"vs_gpt4o_saving":"85%","channels":[{"input_per_million_tokens":0.75,"output_per_million_tokens":3.75,"cache_read_input_per_million_tokens":0.075}],"discount_multiplier":1,"discount_source":"none"},{"alias":"gpt-5.4-mini","display_name":"GPT-5.4 Mini","description":"Fast, efficient model balancing capability and cost for high-volume workloads.","category":"chat","categories":["chat"],"max_input_tokens":400000,"max_output_tokens":128000,"capabilities":["chat","vision","code","tools","reasoning"],"input_modalities":["text","image","file"],"supports_structured_output":true,"is_chinese_model":false,"vs_gpt4o_saving":"85%","channels":[{"input_per_million_tokens":0.75,"output_per_million_tokens":4.5,"cache_read_input_per_million_tokens":0.375},{"input_per_million_tokens":0.75,"output_per_million_tokens":4.5,"cache_read_input_per_million_tokens":0.375}],"discount_multiplier":1,"discount_source":"none"},{"alias":"gemini-3.1-flash-live","display_name":"gemini-3.1-flash-live","description":"Google Gemini 3.1 Flash Live speech-to-speech over WebSocket (AI Studio BidiGenerateContent). Flagship low-latency voice conversation. Audio $3/M in · $12/M out; text $0.75/M in · $4.50/M out. Billed per response at official list price.","category":"realtime","categories":["realtime"],"capabilities":["realtime","audio"],"input_modalities":["audio","text"],"supports_structured_output":false,"is_chinese_model":false,"vs_gpt4o_saving":"85%","channels":[{"input_per_million_tokens":0.75,"output_per_million_tokens":4.5}],"release_stage":"invited_beta","discount_multiplier":1,"discount_source":"none"},{"alias":"kimi-k2.7-code","display_name":"Kimi K2.7 Code","description":"Moonshot's strongest coding \u0026 agentic model.","category":"chat","categories":["chat"],"max_input_tokens":256000,"max_output_tokens":131072,"capabilities":["chat","code","reasoning","tools"],"input_modalities":["text"],"supports_structured_output":true,"is_chinese_model":true,"vs_gpt4o_saving":"81%","channels":[{"input_per_million_tokens":0.95,"output_per_million_tokens":4,"cache_read_input_per_million_tokens":0.19},{"input_per_million_tokens":0.95,"output_per_million_tokens":4,"cache_read_input_per_million_tokens":0.19},{"input_per_million_tokens":0.95,"output_per_million_tokens":4,"cache_read_input_per_million_tokens":0.19}],"discount_multiplier":1,"discount_source":"none"},{"alias":"glm-5","display_name":"GLM-5","description":"Z.ai's 5th-gen flagship foundation model; thinking + tools, 200K context.","category":"chat","categories":["chat"],"max_input_tokens":200000,"max_output_tokens":131072,"capabilities":["chat","code","reasoning","tools","long-context"],"input_modalities":["text"],"supports_structured_output":true,"is_chinese_model":true,"vs_gpt4o_saving":"80%","channels":[{"input_per_million_tokens":1,"output_per_million_tokens":3.2,"cache_read_input_per_million_tokens":0.2},{"input_per_million_tokens":1,"output_per_million_tokens":3.2,"cache_read_input_per_million_tokens":0.2},{"input_per_million_tokens":1,"output_per_million_tokens":3.2,"cache_read_input_per_million_tokens":0.2}],"discount_multiplier":1,"discount_source":"none"},{"alias":"claude-haiku-4-5","display_name":"Claude Haiku 4.5","description":"Lightweight, high-efficiency model optimized for speed and cost.","category":"chat","categories":["chat"],"max_input_tokens":200000,"max_output_tokens":64000,"capabilities":["chat","code","tools"],"input_modalities":["text","image","file"],"supports_structured_output":true,"is_chinese_model":false,"vs_gpt4o_saving":"80%","channels":[{"input_per_million_tokens":1,"output_per_million_tokens":5,"cache_read_input_per_million_tokens":0.1,"cache_write_5m_input_per_million_tokens":1.25,"cache_write_1h_input_per_million_tokens":2},{"input_per_million_tokens":1,"output_per_million_tokens":5,"cache_read_input_per_million_tokens":0.1,"cache_write_5m_input_per_million_tokens":1.25,"cache_write_1h_input_per_million_tokens":2},{"input_per_million_tokens":1,"output_per_million_tokens":5,"cache_read_input_per_million_tokens":0.1,"cache_write_5m_input_per_million_tokens":1.25,"cache_write_1h_input_per_million_tokens":2},{"input_per_million_tokens":1,"output_per_million_tokens":5,"cache_read_input_per_million_tokens":0.1,"cache_write_5m_input_per_million_tokens":1.25,"cache_write_1h_input_per_million_tokens":2},{"input_per_million_tokens":1,"output_per_million_tokens":5,"cache_read_input_per_million_tokens":0.1,"cache_write_5m_input_per_million_tokens":1.25,"cache_write_1h_input_per_million_tokens":2},{"input_per_million_tokens":1,"output_per_million_tokens":5,"cache_read_input_per_million_tokens":0.1,"cache_write_5m_input_per_million_tokens":1.25,"cache_write_1h_input_per_million_tokens":2},{"input_per_million_tokens":1,"output_per_million_tokens":5,"cache_read_input_per_million_tokens":0.1,"cache_write_5m_input_per_million_tokens":1.25,"cache_write_1h_input_per_million_tokens":2}],"discount_multiplier":1,"discount_source":"none"},{"alias":"gpt-5.6-luna","display_name":"GPT-5.6 Luna","description":"Fastest, most cost-efficient GPT-5.6 tier (2026-07).","category":"chat","categories":["chat"],"max_input_tokens":1050000,"max_output_tokens":128000,"capabilities":["chat","vision","code","tools","reasoning"],"input_modalities":["text","image","file"],"supports_structured_output":true,"is_chinese_model":false,"vs_gpt4o_saving":"80%","channels":[{"input_per_million_tokens":1,"output_per_million_tokens":6,"cache_read_input_per_million_tokens":0.1,"cache_write_input_per_million_tokens":1.25},{"input_per_million_tokens":1,"output_per_million_tokens":6,"cache_read_input_per_million_tokens":0.1,"cache_write_input_per_million_tokens":1.25},{"input_per_million_tokens":1,"output_per_million_tokens":6,"cache_read_input_per_million_tokens":0.1,"cache_write_input_per_million_tokens":1.25},{"input_per_million_tokens":1,"output_per_million_tokens":6,"cache_read_input_per_million_tokens":0.1,"cache_write_input_per_million_tokens":1.25}],"discount_multiplier":1,"discount_source":"none"},{"alias":"qwen3-max","display_name":"Qwen3 Max","description":"Qwen's strongest reasoning model.","category":"chat","categories":["chat"],"max_input_tokens":256000,"max_output_tokens":16384,"capabilities":["chat","code","reasoning","tools"],"input_modalities":["text"],"supports_structured_output":true,"is_chinese_model":true,"vs_gpt4o_saving":"76%","channels":[{"input_per_million_tokens":1.2,"output_per_million_tokens":6,"cache_read_input_per_million_tokens":0.359,"tiers":[{"max_input_tokens":32000,"input_per_m":1.2,"output_per_m":6},{"max_input_tokens":128000,"input_per_m":2.4,"output_per_m":12},{"max_input_tokens":256000,"input_per_m":3,"output_per_m":15}]}],"discount_multiplier":1,"discount_source":"none"},{"alias":"gpt-4o-mini-transcribe","display_name":"GPT-4o Mini Transcribe","description":"Cost-efficient token-based transcription. $1.25/M input, $5/M output.","category":"audio","categories":["audio"],"max_input_tokens":16000,"capabilities":["transcription"],"input_modalities":["audio"],"supports_structured_output":false,"is_chinese_model":false,"vs_gpt4o_saving":"75%","channels":[{"input_per_million_tokens":1.25,"output_per_million_tokens":5}],"discount_multiplier":1,"discount_source":"none"},{"alias":"gemini-2.5-pro","display_name":"Gemini 2.5 Pro","description":"Powerful general-purpose model with long context and thinking support.","category":"chat","categories":["chat"],"max_input_tokens":1048576,"max_output_tokens":65536,"capabilities":["chat","vision","code","tools","reasoning","long-context"],"input_modalities":["text","image","audio","video","file"],"supports_structured_output":true,"is_chinese_model":false,"vs_gpt4o_saving":"75%","channels":[{"input_per_million_tokens":1.25,"output_per_million_tokens":10,"cache_read_input_per_million_tokens":0.125},{"input_per_million_tokens":1.25,"output_per_million_tokens":10,"cache_read_input_per_million_tokens":0.125}],"discount_multiplier":1,"discount_source":"none"},{"alias":"deepseek-v4-pro","display_name":"DeepSeek V4 Pro","description":"DeepSeek V4 Pro. Higher quality reasoning, slower + more expensive than Flash.","category":"chat","categories":["chat"],"max_input_tokens":1000000,"max_output_tokens":393216,"capabilities":["chat","code","reasoning","tools"],"input_modalities":["text"],"supports_structured_output":true,"is_chinese_model":true,"vs_gpt4o_saving":"74%","channels":[{"input_per_million_tokens":1.32,"output_per_million_tokens":3.96,"cache_read_input_per_million_tokens":0.132},{"input_per_million_tokens":1.32,"output_per_million_tokens":3.96,"cache_read_input_per_million_tokens":0.132}],"discount_multiplier":1,"discount_source":"none"},{"alias":"deepseek-v4-pro-0813","display_name":"DeepSeek V4 Pro (0813)","description":"DeepSeek V4 Pro 2026-08-13 snapshot. 1M context, reasoning-heavy, five reasoning_effort levels.","category":"chat","categories":["chat"],"max_input_tokens":1000000,"max_output_tokens":393216,"capabilities":["chat","code","reasoning","tools"],"input_modalities":["text"],"supports_structured_output":true,"is_chinese_model":true,"vs_gpt4o_saving":"74%","channels":[{"input_per_million_tokens":1.32,"output_per_million_tokens":3.96,"cache_read_input_per_million_tokens":0.132}],"discount_multiplier":1,"discount_source":"none"},{"alias":"glm-5.2","display_name":"GLM-5.2","description":"Z.ai's flagship GLM-5.2 reasoning \u0026 coding model with thinking. 1M-token context.","category":"chat","categories":["chat"],"max_input_tokens":1048576,"max_output_tokens":131072,"capabilities":["chat","code","reasoning","tools","long-context"],"input_modalities":["text"],"supports_structured_output":true,"is_chinese_model":true,"vs_gpt4o_saving":"72%","channels":[{"input_per_million_tokens":1.4,"output_per_million_tokens":4.4,"cache_read_input_per_million_tokens":0.26},{"input_per_million_tokens":1.4,"output_per_million_tokens":4.4,"cache_read_input_per_million_tokens":0.26},{"input_per_million_tokens":1.4,"output_per_million_tokens":4.4,"cache_read_input_per_million_tokens":0.26},{"input_per_million_tokens":1.4,"output_per_million_tokens":4.4,"cache_read_input_per_million_tokens":0.28},{"input_per_million_tokens":1.4,"output_per_million_tokens":4.4,"cache_read_input_per_million_tokens":0.28}],"discount_multiplier":1,"discount_source":"none"},{"alias":"glm-5.3","display_name":"GLM-5.3","description":"Z.ai's GLM-5.3 reasoning \u0026 coding model for long-horizon agent work. Always-on thinking, 1M-token context.","category":"chat","categories":["chat"],"max_input_tokens":1048576,"max_output_tokens":131072,"capabilities":["chat","code","reasoning","tools","long-context"],"input_modalities":["text"],"supports_structured_output":true,"is_chinese_model":true,"vs_gpt4o_saving":"72%","channels":[{"input_per_million_tokens":1.4,"output_per_million_tokens":4.4,"cache_read_input_per_million_tokens":0.26}],"discount_multiplier":1,"discount_source":"none"},{"alias":"gemini-3.6-flash","display_name":"Gemini 3.6 Flash","description":"Faster, cheaper Gemini 3.6 Flash (output $7.50/M, down from 3.5's $9).","category":"chat","categories":["chat"],"max_input_tokens":1048576,"max_output_tokens":65536,"capabilities":["chat","vision","code","tools"],"input_modalities":["text","image","audio","video","file"],"supports_structured_output":true,"is_chinese_model":false,"vs_gpt4o_saving":"70%","channels":[{"input_per_million_tokens":1.5,"output_per_million_tokens":7.5,"cache_read_input_per_million_tokens":0.15}],"discount_multiplier":1,"discount_source":"none"},{"alias":"gemini-3.5-flash","display_name":"Gemini 3.5 Flash","description":"Fast, multimodal model with audio input support. $1.50/M input, $9/M output.","category":"chat","categories":["chat"],"max_input_tokens":1048576,"max_output_tokens":65536,"capabilities":["chat","vision","code","tools"],"input_modalities":["text","image","audio","video","file"],"supports_structured_output":true,"is_chinese_model":false,"vs_gpt4o_saving":"70%","channels":[{"input_per_million_tokens":1.5,"output_per_million_tokens":9,"cache_read_input_per_million_tokens":0.15},{"input_per_million_tokens":1.5,"output_per_million_tokens":9,"cache_read_input_per_million_tokens":0.15}],"discount_multiplier":1,"discount_source":"none"},{"alias":"gpt-5.2","display_name":"GPT-5.2","description":"Prior-generation flagship. Serves on both Chat Completions and Responses.","category":"chat","categories":["chat"],"max_input_tokens":400000,"max_output_tokens":128000,"capabilities":["chat","vision","code","tools","reasoning"],"input_modalities":["text","image","file"],"supports_structured_output":true,"is_chinese_model":false,"vs_gpt4o_saving":"65%","channels":[{"input_per_million_tokens":1.75,"output_per_million_tokens":14,"cache_read_input_per_million_tokens":0.875}],"discount_multiplier":1,"discount_source":"none"},{"alias":"gpt-5.3-codex","display_name":"GPT-5.3 Codex","description":"Specialised coding / tool-use model; OpenAI Responses API only.","category":"chat","categories":["chat"],"max_input_tokens":400000,"max_output_tokens":128000,"capabilities":["chat","code","tools","reasoning"],"input_modalities":["text","image","file"],"supports_structured_output":true,"is_chinese_model":false,"vs_gpt4o_saving":"65%","channels":[{"input_per_million_tokens":1.75,"output_per_million_tokens":14,"cache_read_input_per_million_tokens":0.875}],"discount_multiplier":1,"discount_source":"none"},{"alias":"qwen3.8-max","display_name":"Qwen3.8 Max","description":"Alibaba's flagship Qwen3.8 Max — 2.4T-param MoE, natively vision-language, strong reasoning \u0026 coding.","category":"chat","categories":["chat"],"max_input_tokens":983616,"max_output_tokens":131072,"capabilities":["chat","vision","code","reasoning","tools","long-context"],"input_modalities":["text","image"],"supports_structured_output":true,"is_chinese_model":true,"vs_gpt4o_saving":"60%","channels":[{"input_per_million_tokens":2,"output_per_million_tokens":6,"cache_read_input_per_million_tokens":0.25,"cache_write_input_per_million_tokens":2.5,"explicit_cache_read_input_per_million_tokens":0.17}],"discount_multiplier":1,"discount_source":"none"},{"alias":"claude-sonnet-5","display_name":"Claude Sonnet 5","description":"Anthropic's latest Sonnet-tier model; best combination of speed and intelligence, adaptive thinking, doubled max output vs Sonnet 4.6.","category":"chat","categories":["chat"],"max_input_tokens":1000000,"max_output_tokens":128000,"capabilities":["chat","code","thinking","tools","reasoning"],"input_modalities":["text","image","file"],"supports_structured_output":true,"is_chinese_model":false,"vs_gpt4o_saving":"60%","channels":[{"input_per_million_tokens":2,"output_per_million_tokens":10,"cache_read_input_per_million_tokens":0.2,"cache_write_5m_input_per_million_tokens":2.5,"cache_write_1h_input_per_million_tokens":4},{"input_per_million_tokens":2,"output_per_million_tokens":10,"cache_read_input_per_million_tokens":0.2,"cache_write_5m_input_per_million_tokens":2.5,"cache_write_1h_input_per_million_tokens":4},{"input_per_million_tokens":2,"output_per_million_tokens":10,"cache_read_input_per_million_tokens":0.2,"cache_write_5m_input_per_million_tokens":2.5,"cache_write_1h_input_per_million_tokens":4},{"input_per_million_tokens":2,"output_per_million_tokens":10,"cache_read_input_per_million_tokens":0.2,"cache_write_5m_input_per_million_tokens":2.5,"cache_write_1h_input_per_million_tokens":4}],"discount_multiplier":1,"discount_source":"none"},{"alias":"gemini-3-pro-image-preview","display_name":"gemini-3-pro-image-preview","description":"","category":"image","categories":["image"],"capabilities":["image"],"input_modalities":null,"supports_structured_output":false,"is_chinese_model":false,"vs_gpt4o_saving":"60%","channels":[{"input_per_million_tokens":2,"output_per_million_tokens":120}],"discount_multiplier":1,"discount_source":"none"},{"alias":"gemini-3.1-pro-preview","display_name":"Gemini 3.1 Pro","description":"Google's latest frontier reasoning model with thinking capabilities.","category":"chat","categories":["chat"],"max_input_tokens":1048576,"max_output_tokens":65536,"capabilities":["chat","vision","code","tools","reasoning","image"],"input_modalities":["text","image","audio","video","file"],"supports_structured_output":true,"is_chinese_model":false,"vs_gpt4o_saving":"60%","channels":[{"input_per_million_tokens":2,"output_per_million_tokens":12,"cache_read_input_per_million_tokens":0.2},{"input_per_million_tokens":2,"output_per_million_tokens":12,"cache_read_input_per_million_tokens":0.2}],"discount_multiplier":1,"discount_source":"none"},{"alias":"gpt-image-1-mini","display_name":"gpt-image-1-mini","description":"","category":"image","categories":["image"],"capabilities":["image"],"input_modalities":null,"supports_structured_output":false,"is_chinese_model":false,"vs_gpt4o_saving":"60%","channels":[{"input_per_million_tokens":2,"output_per_million_tokens":8}],"discount_multiplier":1,"discount_source":"none"},{"alias":"gpt-5.4","display_name":"GPT-5.4","description":"OpenAI's frontier model for complex reasoning, coding, and multimodal tasks.","category":"chat","categories":["chat"],"max_input_tokens":922000,"max_output_tokens":128000,"capabilities":["chat","vision","code","tools","reasoning"],"input_modalities":["text","image","file"],"supports_structured_output":true,"is_chinese_model":false,"vs_gpt4o_saving":"50%","channels":[{"input_per_million_tokens":2.5,"output_per_million_tokens":15,"cache_read_input_per_million_tokens":1.25},{"input_per_million_tokens":2.5,"output_per_million_tokens":15,"cache_read_input_per_million_tokens":1.25}],"discount_multiplier":1,"discount_source":"none"},{"alias":"gpt-5.6-terra","display_name":"GPT-5.6 Terra","description":"Balanced mid-tier of the GPT-5.6 family (2026-07).","category":"chat","categories":["chat"],"max_input_tokens":1050000,"max_output_tokens":128000,"capabilities":["chat","vision","code","tools","reasoning"],"input_modalities":["text","image","file"],"supports_structured_output":true,"is_chinese_model":false,"vs_gpt4o_saving":"50%","channels":[{"input_per_million_tokens":2.5,"output_per_million_tokens":15,"cache_read_input_per_million_tokens":0.25,"cache_write_input_per_million_tokens":3.125},{"input_per_million_tokens":2.5,"output_per_million_tokens":15,"cache_read_input_per_million_tokens":0.25,"cache_write_input_per_million_tokens":3.125},{"input_per_million_tokens":2.5,"output_per_million_tokens":15,"cache_read_input_per_million_tokens":0.25,"cache_write_input_per_million_tokens":3.125},{"input_per_million_tokens":2.5,"output_per_million_tokens":15,"cache_read_input_per_million_tokens":0.25,"cache_write_input_per_million_tokens":3.125}],"discount_multiplier":1,"discount_source":"none"},{"alias":"gpt-4o-transcribe","display_name":"GPT-4o Transcribe","description":"Token-based transcription with streaming and logprobs. $2.50/M input, $10/M output.","category":"audio","categories":["audio"],"max_input_tokens":16000,"capabilities":["transcription"],"input_modalities":["audio"],"supports_structured_output":false,"is_chinese_model":false,"vs_gpt4o_saving":"50%","channels":[{"input_per_million_tokens":2.5,"output_per_million_tokens":10}],"discount_multiplier":1,"discount_source":"none"},{"alias":"qwen3.7-max","display_name":"Qwen3.7 Max","description":"Alibaba's flagship Qwen3.7 reasoning \u0026 coding model. 1M-token context.","category":"chat","categories":["chat"],"max_input_tokens":1000000,"max_output_tokens":65536,"capabilities":["chat","code","reasoning","tools","long-context"],"input_modalities":["text"],"supports_structured_output":true,"is_chinese_model":true,"vs_gpt4o_saving":"50%","channels":[{"input_per_million_tokens":2.5,"output_per_million_tokens":7.5,"cache_read_input_per_million_tokens":0.5},{"input_per_million_tokens":2.5,"output_per_million_tokens":7.5,"cache_read_input_per_million_tokens":0.5}],"discount_multiplier":1,"discount_source":"none"},{"alias":"gpt-4o-transcribe-diarize","display_name":"GPT-4o Transcribe Diarize","description":"Token-based transcription with speaker diarization. Supports diarized_json response format with per-segment speaker labels. audio-in $6.25/M, out ~$2.50/M (dashboard-calibrated).","category":"audio","categories":["audio"],"max_input_tokens":16000,"capabilities":["transcription"],"input_modalities":["audio"],"supports_structured_output":false,"is_chinese_model":false,"vs_gpt4o_saving":"50%","channels":[{"input_per_million_tokens":2.5,"output_per_million_tokens":2.5}],"discount_multiplier":1,"discount_source":"none"},{"alias":"claude-sonnet-4","display_name":"Claude Sonnet 4","description":"Anthropic Sonnet 4: balanced model for coding and everyday agentic tasks.","category":"chat","categories":["chat"],"max_input_tokens":200000,"max_output_tokens":64000,"capabilities":["chat","code","thinking","tools","reasoning"],"input_modalities":["text","image"],"supports_structured_output":true,"is_chinese_model":false,"vs_gpt4o_saving":"40%","channels":[{"input_per_million_tokens":3,"output_per_million_tokens":15,"cache_read_input_per_million_tokens":0.3,"cache_write_5m_input_per_million_tokens":3.75,"cache_write_1h_input_per_million_tokens":6}],"release_stage":"invited_beta","discount_multiplier":1,"discount_source":"none"},{"alias":"claude-sonnet-4-5","display_name":"Claude Sonnet 4.5","description":"Balanced Claude 4.5 model. Good mix of intelligence, speed, and cost.","category":"chat","categories":["chat"],"max_input_tokens":200000,"max_output_tokens":64000,"capabilities":["chat","code","thinking","tools","reasoning"],"input_modalities":["text","image","file"],"supports_structured_output":true,"is_chinese_model":false,"vs_gpt4o_saving":"40%","channels":[{"input_per_million_tokens":3,"output_per_million_tokens":15,"cache_read_input_per_million_tokens":0.3,"cache_write_5m_input_per_million_tokens":3.75,"cache_write_1h_input_per_million_tokens":6},{"input_per_million_tokens":3,"output_per_million_tokens":15,"cache_read_input_per_million_tokens":0.3,"cache_write_5m_input_per_million_tokens":3.75,"cache_write_1h_input_per_million_tokens":6},{"input_per_million_tokens":3,"output_per_million_tokens":15,"cache_read_input_per_million_tokens":0.3,"cache_write_5m_input_per_million_tokens":3.75,"cache_write_1h_input_per_million_tokens":6},{"input_per_million_tokens":3,"output_per_million_tokens":15,"cache_read_input_per_million_tokens":0.3,"cache_write_5m_input_per_million_tokens":3.75,"cache_write_1h_input_per_million_tokens":6},{"input_per_million_tokens":3,"output_per_million_tokens":15,"cache_read_input_per_million_tokens":0.3,"cache_write_5m_input_per_million_tokens":3.75,"cache_write_1h_input_per_million_tokens":6}],"discount_multiplier":1,"discount_source":"none"},{"alias":"claude-sonnet-4-6","display_name":"Claude Sonnet 4.6","description":"High-performance model balancing frontier intelligence, speed, and cost. Top choice for computer use and agentic workflows.","category":"chat","categories":["chat"],"max_input_tokens":1000000,"max_output_tokens":64000,"capabilities":["chat","code","thinking","tools","reasoning"],"input_modalities":["text","image","file"],"supports_structured_output":true,"is_chinese_model":false,"vs_gpt4o_saving":"40%","channels":[{"input_per_million_tokens":3,"output_per_million_tokens":15,"cache_read_input_per_million_tokens":0.3,"cache_write_5m_input_per_million_tokens":3.75,"cache_write_1h_input_per_million_tokens":6},{"input_per_million_tokens":3,"output_per_million_tokens":15,"cache_read_input_per_million_tokens":0.3,"cache_write_5m_input_per_million_tokens":3.75,"cache_write_1h_input_per_million_tokens":6},{"input_per_million_tokens":3,"output_per_million_tokens":15,"cache_read_input_per_million_tokens":0.3,"cache_write_5m_input_per_million_tokens":3.75,"cache_write_1h_input_per_million_tokens":6},{"input_per_million_tokens":3,"output_per_million_tokens":15,"cache_read_input_per_million_tokens":0.3,"cache_write_5m_input_per_million_tokens":3.75,"cache_write_1h_input_per_million_tokens":6},{"input_per_million_tokens":3,"output_per_million_tokens":15,"cache_read_input_per_million_tokens":0.3,"cache_write_5m_input_per_million_tokens":3.75,"cache_write_1h_input_per_million_tokens":6},{"input_per_million_tokens":3,"output_per_million_tokens":15,"cache_read_input_per_million_tokens":0.3,"cache_write_5m_input_per_million_tokens":3.75,"cache_write_1h_input_per_million_tokens":6},{"input_per_million_tokens":3,"output_per_million_tokens":15,"cache_read_input_per_million_tokens":0.3,"cache_write_5m_input_per_million_tokens":3.75,"cache_write_1h_input_per_million_tokens":6},{"input_per_million_tokens":3,"output_per_million_tokens":15,"cache_read_input_per_million_tokens":0.3,"cache_write_5m_input_per_million_tokens":3.75,"cache_write_1h_input_per_million_tokens":6}],"discount_multiplier":1,"discount_source":"none"},{"alias":"kimi-k3","display_name":"Kimi K3","description":"Moonshot's flagship K3 reasoning model; thinking + tools, 1M-token context.","category":"chat","categories":["chat"],"max_input_tokens":1000000,"max_output_tokens":131072,"capabilities":["chat","code","reasoning","tools"],"input_modalities":["text"],"supports_structured_output":true,"is_chinese_model":true,"vs_gpt4o_saving":"40%","channels":[{"input_per_million_tokens":3,"output_per_million_tokens":15,"cache_read_input_per_million_tokens":0.3},{"input_per_million_tokens":3,"output_per_million_tokens":15,"cache_read_input_per_million_tokens":0.3},{"input_per_million_tokens":3,"output_per_million_tokens":15,"cache_read_input_per_million_tokens":0.3}],"discount_multiplier":1,"discount_source":"none"},{"alias":"gpt-realtime","display_name":"GPT Realtime","description":"OpenAI speech-to-speech realtime model over WebSocket (/v1/realtime). Audio $32/M in · $64/M out; text $4/M in · $16/M out; cached input $0.40/M. Billed per response at official list price.","category":"realtime","categories":["realtime"],"capabilities":["realtime","audio"],"input_modalities":["audio","text"],"supports_structured_output":false,"is_chinese_model":false,"vs_gpt4o_saving":"20%","channels":[{"input_per_million_tokens":4,"output_per_million_tokens":16,"cache_read_input_per_million_tokens":0.4}],"release_stage":"invited_beta","discount_multiplier":1,"discount_source":"none"},{"alias":"google-tts-standard","display_name":"Google TTS Standard","description":"Google Cloud Text-to-Speech, Standard voices. The cheapest tier on the gateway — billed per input character at $4 per 1M characters.","category":"other","categories":["other"],"max_input_tokens":4096,"capabilities":["speech"],"input_modalities":["text"],"supports_structured_output":false,"is_chinese_model":false,"vs_gpt4o_saving":"20%","channels":[{"input_per_million_tokens":4,"output_per_million_tokens":0}],"discount_multiplier":1,"discount_source":"none"},{"alias":"gpt-realtime-2.1","display_name":"GPT Realtime 2.1","description":"OpenAI flagship speech-to-speech realtime model over WebSocket (/v1/realtime). Audio $32/M in · $64/M out; text $4/M in · $24/M out; cached input $0.40/M. Billed per response at official list price.","category":"realtime","categories":["realtime"],"capabilities":["realtime","audio"],"input_modalities":["audio","text"],"supports_structured_output":false,"is_chinese_model":false,"vs_gpt4o_saving":"20%","channels":[{"input_per_million_tokens":4,"output_per_million_tokens":24,"cache_read_input_per_million_tokens":0.4},{"input_per_million_tokens":4,"output_per_million_tokens":24,"cache_read_input_per_million_tokens":0.4}],"release_stage":"invited_beta","discount_multiplier":1,"discount_source":"none"},{"alias":"gpt-5.6-sol","display_name":"GPT-5.6 Sol","description":"OpenAI's flagship (2026-07). 1M context, vision, tools, reasoning.","category":"chat","categories":["chat"],"max_input_tokens":1050000,"max_output_tokens":128000,"capabilities":["chat","vision","code","tools","reasoning"],"input_modalities":["text","image","file"],"supports_structured_output":true,"is_chinese_model":false,"channels":[{"input_per_million_tokens":5,"output_per_million_tokens":30,"cache_read_input_per_million_tokens":0.5,"cache_write_input_per_million_tokens":6.25},{"input_per_million_tokens":5,"output_per_million_tokens":30,"cache_read_input_per_million_tokens":0.5,"cache_write_input_per_million_tokens":6.25},{"input_per_million_tokens":5,"output_per_million_tokens":30,"cache_read_input_per_million_tokens":0.5,"cache_write_input_per_million_tokens":6.25},{"input_per_million_tokens":5,"output_per_million_tokens":30,"cache_read_input_per_million_tokens":0.5,"cache_write_input_per_million_tokens":6.25}],"discount_multiplier":1,"discount_source":"none"},{"alias":"gpt-5.6","display_name":"GPT-5.6","description":"Alias of GPT-5.6 Sol (official default routing).","category":"chat","categories":["chat"],"max_input_tokens":1050000,"max_output_tokens":128000,"capabilities":["chat","vision","code","tools","reasoning"],"input_modalities":["text","image","file"],"supports_structured_output":true,"is_chinese_model":false,"channels":[{"input_per_million_tokens":5,"output_per_million_tokens":30,"cache_read_input_per_million_tokens":0.5,"cache_write_input_per_million_tokens":6.25},{"input_per_million_tokens":5,"output_per_million_tokens":30,"cache_read_input_per_million_tokens":0.5,"cache_write_input_per_million_tokens":6.25}],"discount_multiplier":1,"discount_source":"none"},{"alias":"claude-opus-4-5","display_name":"Claude Opus 4.5","description":"Anthropic's flagship Claude 4.5 model. Strong reasoning, coding, and agentic task performance.","category":"chat","categories":["chat"],"max_input_tokens":200000,"max_output_tokens":64000,"capabilities":["chat","code","thinking","tools","reasoning"],"input_modalities":["text","image","file"],"supports_structured_output":true,"is_chinese_model":false,"channels":[{"input_per_million_tokens":5,"output_per_million_tokens":25,"cache_read_input_per_million_tokens":0.5,"cache_write_5m_input_per_million_tokens":6.25,"cache_write_1h_input_per_million_tokens":10},{"input_per_million_tokens":5,"output_per_million_tokens":25,"cache_read_input_per_million_tokens":0.5,"cache_write_5m_input_per_million_tokens":6.25,"cache_write_1h_input_per_million_tokens":10},{"input_per_million_tokens":5,"output_per_million_tokens":25,"cache_read_input_per_million_tokens":0.5,"cache_write_5m_input_per_million_tokens":6.25,"cache_write_1h_input_per_million_tokens":10}],"discount_multiplier":1,"discount_source":"none"},{"alias":"gpt-image-1.5","display_name":"gpt-image-1.5","description":"","category":"image","categories":["image"],"capabilities":["image"],"input_modalities":null,"supports_structured_output":false,"is_chinese_model":false,"channels":[{"input_per_million_tokens":5,"output_per_million_tokens":32}],"discount_multiplier":1,"discount_source":"none"},{"alias":"gpt-image-1","display_name":"gpt-image-1","description":"","category":"image","categories":["image"],"capabilities":["image"],"input_modalities":null,"supports_structured_output":false,"is_chinese_model":false,"channels":[{"input_per_million_tokens":5,"output_per_million_tokens":40}],"discount_multiplier":1,"discount_source":"none"},{"alias":"gpt-5.5","display_name":"GPT-5.5","description":"OpenAI's frontier model (2026-04). 1M context, vision, tools, reasoning.","category":"chat","categories":["chat"],"max_input_tokens":1050000,"max_output_tokens":128000,"capabilities":["chat","vision","code","tools","reasoning"],"input_modalities":["text","image","file"],"supports_structured_output":true,"is_chinese_model":false,"channels":[{"input_per_million_tokens":5,"output_per_million_tokens":30,"cache_read_input_per_million_tokens":0.5},{"input_per_million_tokens":5,"output_per_million_tokens":30,"cache_read_input_per_million_tokens":0.5},{"input_per_million_tokens":5,"output_per_million_tokens":30,"cache_read_input_per_million_tokens":0.5}],"discount_multiplier":1,"discount_source":"none"},{"alias":"claude-opus-4-6","display_name":"Claude Opus 4.6","description":"Anthropic's most intelligent model. Excels at complex coding, multi-step enterprise workflows, and autonomous agentic tasks.","category":"chat","categories":["chat"],"max_input_tokens":1000000,"max_output_tokens":128000,"capabilities":["chat","code","thinking","tools","reasoning"],"input_modalities":["text","image","file"],"supports_structured_output":true,"is_chinese_model":false,"channels":[{"input_per_million_tokens":5,"output_per_million_tokens":25,"cache_read_input_per_million_tokens":0.5,"cache_write_5m_input_per_million_tokens":6.25,"cache_write_1h_input_per_million_tokens":10},{"input_per_million_tokens":5,"output_per_million_tokens":25,"cache_read_input_per_million_tokens":0.5,"cache_write_5m_input_per_million_tokens":6.25,"cache_write_1h_input_per_million_tokens":10},{"input_per_million_tokens":5,"output_per_million_tokens":25,"cache_read_input_per_million_tokens":0.5,"cache_write_5m_input_per_million_tokens":6.25,"cache_write_1h_input_per_million_tokens":10},{"input_per_million_tokens":5,"output_per_million_tokens":25,"cache_read_input_per_million_tokens":0.5,"cache_write_5m_input_per_million_tokens":6.25,"cache_write_1h_input_per_million_tokens":10},{"input_per_million_tokens":5,"output_per_million_tokens":25,"cache_read_input_per_million_tokens":0.5,"cache_write_5m_input_per_million_tokens":6.25,"cache_write_1h_input_per_million_tokens":10},{"input_per_million_tokens":5,"output_per_million_tokens":25,"cache_read_input_per_million_tokens":0.5,"cache_write_5m_input_per_million_tokens":6.25,"cache_write_1h_input_per_million_tokens":10}],"discount_multiplier":1,"discount_source":"none"},{"alias":"claude-opus-4-7","display_name":"Claude Opus 4.7","description":"Anthropic's latest flagship; incremental improvement over 4.6 with stronger agentic / coding performance.","category":"chat","categories":["chat"],"max_input_tokens":1000000,"max_output_tokens":128000,"capabilities":["chat","code","thinking","tools","reasoning"],"input_modalities":["text","image","file"],"supports_structured_output":true,"is_chinese_model":false,"channels":[{"input_per_million_tokens":5,"output_per_million_tokens":25,"cache_read_input_per_million_tokens":0.5,"cache_write_5m_input_per_million_tokens":6.25,"cache_write_1h_input_per_million_tokens":10},{"input_per_million_tokens":5,"output_per_million_tokens":25,"cache_read_input_per_million_tokens":0.5,"cache_write_5m_input_per_million_tokens":6.25,"cache_write_1h_input_per_million_tokens":10},{"input_per_million_tokens":5,"output_per_million_tokens":25,"cache_read_input_per_million_tokens":0.5,"cache_write_5m_input_per_million_tokens":6.25,"cache_write_1h_input_per_million_tokens":10},{"input_per_million_tokens":5,"output_per_million_tokens":25,"cache_read_input_per_million_tokens":0.5,"cache_write_5m_input_per_million_tokens":6.25,"cache_write_1h_input_per_million_tokens":10},{"input_per_million_tokens":5,"output_per_million_tokens":25,"cache_read_input_per_million_tokens":0.5,"cache_write_5m_input_per_million_tokens":6.25,"cache_write_1h_input_per_million_tokens":10},{"input_per_million_tokens":5,"output_per_million_tokens":25,"cache_read_input_per_million_tokens":0.5,"cache_write_5m_input_per_million_tokens":6.25,"cache_write_1h_input_per_million_tokens":10},{"input_per_million_tokens":5,"output_per_million_tokens":25,"cache_read_input_per_million_tokens":0.5,"cache_write_5m_input_per_million_tokens":6.25,"cache_write_1h_input_per_million_tokens":10}],"discount_multiplier":1,"discount_source":"none"},{"alias":"claude-opus-4-8","display_name":"Claude Opus 4.8","description":"Anthropic's latest flagship Opus model; strongest agentic and coding performance.","category":"chat","categories":["chat"],"max_input_tokens":1000000,"max_output_tokens":128000,"capabilities":["chat","code","thinking","tools","reasoning"],"input_modalities":["text","image","file"],"supports_structured_output":true,"is_chinese_model":false,"channels":[{"input_per_million_tokens":5,"output_per_million_tokens":25,"cache_read_input_per_million_tokens":0.5,"cache_write_5m_input_per_million_tokens":6.25,"cache_write_1h_input_per_million_tokens":10},{"input_per_million_tokens":5,"output_per_million_tokens":25,"cache_read_input_per_million_tokens":0.5,"cache_write_5m_input_per_million_tokens":6.25,"cache_write_1h_input_per_million_tokens":10},{"input_per_million_tokens":5,"output_per_million_tokens":25,"cache_read_input_per_million_tokens":0.5,"cache_write_5m_input_per_million_tokens":6.25,"cache_write_1h_input_per_million_tokens":10},{"input_per_million_tokens":5,"output_per_million_tokens":25,"cache_read_input_per_million_tokens":0.5,"cache_write_5m_input_per_million_tokens":6.25,"cache_write_1h_input_per_million_tokens":10},{"input_per_million_tokens":5,"output_per_million_tokens":25,"cache_read_input_per_million_tokens":0.5,"cache_write_5m_input_per_million_tokens":6.25,"cache_write_1h_input_per_million_tokens":10},{"input_per_million_tokens":5,"output_per_million_tokens":25,"cache_read_input_per_million_tokens":0.5,"cache_write_5m_input_per_million_tokens":6.25,"cache_write_1h_input_per_million_tokens":10},{"input_per_million_tokens":5,"output_per_million_tokens":25,"cache_read_input_per_million_tokens":0.5,"cache_write_5m_input_per_million_tokens":6.25,"cache_write_1h_input_per_million_tokens":10},{"input_per_million_tokens":5,"output_per_million_tokens":25,"cache_read_input_per_million_tokens":0.5,"cache_write_5m_input_per_million_tokens":6.25,"cache_write_1h_input_per_million_tokens":10}],"discount_multiplier":1,"discount_source":"none"},{"alias":"gpt-image-2","display_name":"gpt-image-2","description":"","category":"image","categories":["image"],"capabilities":["image"],"input_modalities":null,"supports_structured_output":false,"is_chinese_model":false,"channels":[{"input_per_million_tokens":5,"output_per_million_tokens":30}],"discount_multiplier":1,"discount_source":"none"},{"alias":"claude-opus-5","display_name":"Claude Opus 5","description":"Anthropic's flagship Opus model for complex agentic coding and enterprise work.","category":"chat","categories":["chat"],"max_input_tokens":1000000,"max_output_tokens":128000,"capabilities":["chat","code","thinking","tools","reasoning"],"input_modalities":["text","image","file"],"supports_structured_output":true,"is_chinese_model":false,"channels":[{"input_per_million_tokens":5,"output_per_million_tokens":25,"cache_read_input_per_million_tokens":0.5,"cache_write_5m_input_per_million_tokens":6.25,"cache_write_1h_input_per_million_tokens":10},{"input_per_million_tokens":5,"output_per_million_tokens":25,"cache_read_input_per_million_tokens":0.5,"cache_write_5m_input_per_million_tokens":6.25,"cache_write_1h_input_per_million_tokens":10},{"input_per_million_tokens":5,"output_per_million_tokens":25,"cache_read_input_per_million_tokens":0.5,"cache_write_5m_input_per_million_tokens":6.25,"cache_write_1h_input_per_million_tokens":10},{"input_per_million_tokens":5,"output_per_million_tokens":25,"cache_read_input_per_million_tokens":0.5,"cache_write_5m_input_per_million_tokens":6.25,"cache_write_1h_input_per_million_tokens":10},{"input_per_million_tokens":5,"output_per_million_tokens":25,"cache_read_input_per_million_tokens":0.5,"cache_write_5m_input_per_million_tokens":6.25,"cache_write_1h_input_per_million_tokens":10},{"input_per_million_tokens":5,"output_per_million_tokens":25,"cache_read_input_per_million_tokens":0.5,"cache_write_5m_input_per_million_tokens":6.25,"cache_write_1h_input_per_million_tokens":10},{"input_per_million_tokens":5,"output_per_million_tokens":25,"cache_read_input_per_million_tokens":0.5,"cache_write_5m_input_per_million_tokens":6.25,"cache_write_1h_input_per_million_tokens":10}],"discount_multiplier":1,"discount_source":"none"},{"alias":"claude-fable-5","display_name":"Claude Fable 5","description":"Anthropic's most intelligent model (Mythos-class). Adaptive thinking; invited beta.","category":"chat","categories":["chat"],"max_input_tokens":200000,"max_output_tokens":4096,"capabilities":["chat","code","thinking","tools","reasoning"],"input_modalities":["text","image"],"supports_structured_output":true,"is_chinese_model":false,"channels":[{"input_per_million_tokens":10,"output_per_million_tokens":50,"cache_read_input_per_million_tokens":1,"cache_write_5m_input_per_million_tokens":12.5,"cache_write_1h_input_per_million_tokens":20},{"input_per_million_tokens":10,"output_per_million_tokens":50,"cache_read_input_per_million_tokens":1,"cache_write_5m_input_per_million_tokens":12.5,"cache_write_1h_input_per_million_tokens":20},{"input_per_million_tokens":10,"output_per_million_tokens":50,"cache_read_input_per_million_tokens":1,"cache_write_5m_input_per_million_tokens":12.5,"cache_write_1h_input_per_million_tokens":20}],"release_stage":"invited_beta","discount_multiplier":1,"discount_source":"none"},{"alias":"qwen3-tts-instruct-flash","display_name":"Qwen3 TTS Instruct Flash","description":"Alibaba Qwen3 text-to-speech with instruction control. Billed per input character at $11.50 per 1M characters. Shares the same API key and channel as Qwen chat and ASR.","category":"other","categories":["other"],"max_input_tokens":4096,"capabilities":["speech"],"input_modalities":["text"],"supports_structured_output":false,"is_chinese_model":true,"channels":[{"input_per_million_tokens":11.5,"output_per_million_tokens":0}],"discount_multiplier":1,"discount_source":"none"},{"alias":"tts-1","display_name":"TTS-1","description":"OpenAI text-to-speech. Billed per input character at $15/1M characters.","category":"other","categories":["other"],"max_input_tokens":4096,"capabilities":["speech"],"input_modalities":["text"],"supports_structured_output":false,"is_chinese_model":false,"channels":[{"input_per_million_tokens":15,"output_per_million_tokens":0}],"discount_multiplier":1,"discount_source":"none"},{"alias":"google-tts-neural2","display_name":"Google TTS Neural2","description":"Google Cloud Text-to-Speech, Neural2 voices — more natural than Standard. Billed per input character at $16 per 1M characters.","category":"other","categories":["other"],"max_input_tokens":4096,"capabilities":["speech"],"input_modalities":["text"],"supports_structured_output":false,"is_chinese_model":false,"channels":[{"input_per_million_tokens":16,"output_per_million_tokens":0}],"discount_multiplier":1,"discount_source":"none"},{"alias":"tts-1-hd","display_name":"TTS-1 HD","description":"OpenAI high-definition text-to-speech. Billed per input character at $30/1M characters.","category":"other","categories":["other"],"max_input_tokens":4096,"capabilities":["speech"],"input_modalities":["text"],"supports_structured_output":false,"is_chinese_model":false,"channels":[{"input_per_million_tokens":30,"output_per_million_tokens":0}],"discount_multiplier":1,"discount_source":"none"},{"alias":"seed-tts-2.0","display_name":"BytePlus Seed TTS 2.0","description":"ByteDance Seed Speech text-to-speech (overseas/BytePlus). Billed per input character at $30/1M characters. Shares the same API key as Seed ASR.","category":"other","categories":["other"],"max_input_tokens":4096,"capabilities":["speech"],"input_modalities":["text"],"supports_structured_output":false,"is_chinese_model":false,"channels":[{"input_per_million_tokens":30,"output_per_million_tokens":0}],"discount_multiplier":1,"discount_source":"none"},{"alias":"google-tts-chirp3-hd","display_name":"Google TTS Chirp 3 HD","description":"Google Cloud Text-to-Speech, Chirp 3 HD voices — the newest, most expressive tier. Billed per input character at $30 per 1M characters.","category":"other","categories":["other"],"max_input_tokens":4096,"capabilities":["speech"],"input_modalities":["text"],"supports_structured_output":false,"is_chinese_model":false,"channels":[{"input_per_million_tokens":30,"output_per_million_tokens":0}],"discount_multiplier":1,"discount_source":"none"},{"alias":"gpt-5.5-pro","display_name":"GPT-5.5 Pro","description":"Higher-reasoning GPT-5.5 (Responses API). Long-running, no streaming.","category":"chat","categories":["chat"],"max_input_tokens":1050000,"max_output_tokens":128000,"capabilities":["chat","vision","code","tools","reasoning"],"input_modalities":["text","image","file"],"supports_structured_output":true,"is_chinese_model":false,"channels":[{"input_per_million_tokens":30,"output_per_million_tokens":180}],"discount_multiplier":1,"discount_source":"none"},{"alias":"gpt-5.4-pro","display_name":"GPT-5.4 Pro","description":"Higher-reasoning GPT-5.4 variant (Responses API).","category":"chat","categories":["chat"],"max_input_tokens":922000,"max_output_tokens":128000,"capabilities":["chat","vision","code","tools","reasoning"],"input_modalities":["text","image","file"],"supports_structured_output":true,"is_chinese_model":false,"channels":[{"input_per_million_tokens":30,"output_per_million_tokens":180,"cache_read_input_per_million_tokens":15}],"discount_multiplier":1,"discount_source":"none"}],"success":true,"total":115}