New Sign up free, 10 calls on us. Up to $1, no card needed.

Qwen3 Max

Released 2025-09-23

chatCodeReasoningTool callingPrompt caching

Qwen3-Max is the Qwen team's trillion-parameter-scale flagship, introduced in the official post "Qwen3-Max: Just Scale" with a 256K-token context.

Input
text $1.2/M
Output
text $6/M
Cache read
$0.359/M
Context
256K
vs GPT-4o
~76% cheaper

Benchmarks

Qwen3 Max
LiveCodeBench Instruct, v6 (25.02-25.05) per Alibaba (Qwen)
no other model measured
69%
SWE-bench Verified Instruct per Alibaba (Qwen)
no other model measured
69.6%
AIME 2025 Instruct per Alibaba (Qwen)
no other model measured
81.6%
SuperGPQA Instruct per Alibaba (Qwen)
no other model measured
65.1%
τ²-Bench Instruct, weighted per Alibaba (Qwen)
no other model measured
74.8%

Vendor-published: Alibaba (Qwen)

Price in context

Where the price sits among 60 comparable models

Input$1.2/M
$0.05 · Qwen3 VL Flash GPT-5.4 Pro · $30
Output$6/M
$0.275 · DeepSeek V4 Flash GPT-5.4 Pro · $180
Cached read$0.359/M
$0.0028 · DeepSeek V4 Flash GPT-5.4 Pro · $15

The bar shows how this model’s price compares with every other model of the same kind on Synthorai. The cheapest and the most expensive are named at each end. These are base rates; batch, region and cache-write discounts are on the pricing page.

Specs & limits

Tokens

Context window (vendor spec) 256,000
Max output (vendor spec) 65,536

Prompt caching

How it caches automatic + explicit
Min prefix 1,024
Lifetime explicit: 5m, reset on hit
Write cost 1.25x

Thinking

Vendor control enable_thinking + thinking_budget
Accepted values enable_thinking true · false; thinking_budget in tokens
Default off; enable_thinking defaults to false, and thinking_budget defaults to the model's maximum chain-of-thought length applied when the request sets nothing
Can be turned off Yes
Thinking behaviour Trace returns in reasoning_content; reasoning_content in the message history is ignored by default, and qwen3-max is not among the models that accept preserve_thinking.
Parameter reasoning_effort
Values minimal · low · medium · high the gateway's parameter surface - the vendor mapping above applies

Model

Modalities text → text
Parameters 1T+ total MoE
  • First trillion-plus-parameter Qwen flagship (closed weights), text-only
  • officially positioned for agentic coding and tool calling
  • hybrid thinking, disabled by default

per Alibaba official docs ↗

Use Qwen3 Max in 30 seconds

OpenAI-compatible: swap the base_url, keep your SDK. POST /v1/chat/completions

from openai import OpenAI

client = OpenAI(
    base_url="https://synthorai.io/v1",
    api_key="sk-syn-...",
)

resp = client.chat.completions.create(
    model="qwen3-max",
    messages=[{"role": "user", "content": "Summarize this diff"}],
    reasoning_effort="medium",
)
print(resp.choices[0].message.content)

About Qwen3 Max

  • Alibaba reports state-of-the-art results across its own evaluations of knowledge, reasoning, coding, instruction following, agent tasks, and multilingual understanding, and its release note singles out an upgrade for agentic coding and tool calling.
  • The line ships in Instruct form alongside a Thinking variant that integrates a code interpreter and adaptive tool use for harder reasoning problems.
  • As the largest model of the Qwen3 series, it targets demanding production workloads.
  • It is a text-in, text-out model (vision lives in the Qwen3-VL siblings) with up to 65,536 output tokens, and Model Studio lists tool calling, structured output, batch inference, and both explicit and implicit context caching.
  • One behavioural detail separates it from every later Qwen generation on this catalog: it is a hybrid thinking model whose reasoning is switched with the enable_thinking parameter and is off by default, where Qwen3.5 and newer arrive with thinking already on.
  • When reasoning is enabled, a thinking_budget parameter caps the spend, and the trace returns separately in reasoning_content, billed as output.
  • Weights for this hosted flagship are not published, and Alibaba now lists it among legacy models, pointing new projects at the Qwen3.5 and Qwen3.6 series.
  • Synthorai offers it through the OpenAI-compatible chat endpoint.

FAQ

Is the Qwen3 Max API free to try?

Yes: new accounts get 10 trial calls and up to $1 in free credit, no card required. At $1.2/M input tokens, that credit alone covers roughly 104 requests of ~8K tokens against Qwen3 Max.

What is Qwen3 Max best at?

Trillion-parameter-scale flagship with 256K context; thinking variant integrates a code interpreter; adaptive tool use for harder reasoning. See the About section for the full picture from the vendor's own release notes.

How much does Qwen3 Max cost?

Qwen3 Max costs $1.2 per million input tokens and $6 per million output tokens on Synthorai. That is the provider's list price, with no platform markup. Cached input tokens bill at $0.359/M.

Does Qwen3 Max support prompt caching?

Yes: automatic caching is on by default, with an explicit mode for guaranteed savings. Cached input tokens bill at $0.359/M vs $1.2/M uncached; prompts need a 1,024-token stable prefix to cache (TTL explicit: 5m, reset on hit). Prompt caching guide →

How do I get access to Qwen3 Max?

Point your existing OpenAI SDK at base_url="https://synthorai.io/v1", set model="qwen3-max", and you're done. One API key covers every model on the gateway.

Related models

Compare

Every value on this page is transcribed from the vendor's own documentation, linked above, and carries the date it was checked. Prices are compared across the catalogue; specification values that vendors define differently are shown with the difference stated rather than charted. Nothing here is measured by us, and nothing is scored.

Get your API key Compare your cost →