ByteDance-Seed-1.8 is a deep-thinking model on BytePlus ModelArk, currently in beta, that its model page credits with stronger multimodal understanding and agent capabilities and superior performance across a wide range of complex real-world tasks.
- Input
- text image video $0.25/M
- Output
- text $2/M
- Cache read
- $0.05/M
- Context
- 262K
- vs GPT-4o
- ~95% cheaper
Price in context
Where the price sits among 60 comparable models
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 |
Thinking
| Vendor control | thinking.type + reasoning_effort |
|---|---|
| Accepted values | thinking.type enabled · disabled (no auto); reasoning_effort minimal · low · medium · high |
| Default | enabled, with reasoning_effort medium applied when the request sets nothing |
| Can be turned off | Yes |
| Thinking behaviour | Trace returns in reasoning_content; from seed-1.8 onward it is kept in the conversation history rather than discarded, and the model decides whether to feed it back into inference. |
| Parameter | reasoning_effort |
| Values | minimal · low · medium · high the gateway's parameter surface - the vendor mapping above applies |
Model
| Modalities | text + image + video → text |
|---|
- Generalized agentic model (search/code/GUI-agent capabilities, native vision)
- ModelArk id seed-1-8 (version seed-1-8-251228, Beta)
- 256K context, 64K max output incl. CoT
- deep reasoning minimal/low/medium/high
- strict-mode function calling
One prompt, measured through the gateway
ByteDance Seed 1.8 passed · 3 sentences
Enabling reasoning capabilities means our LLMs now run multi-step inference chains instead of generating short, direct responses, which doubles the compute resources consumed per query since each step requires processing additional context and intermediate outputs to arrive at a logical conclusion. Additionally, teams across finance, legal, and operations have rapidly adopted these reasoning tools for high-complexity workflows—like automated contract clause analysis and quarterly forecast variance checks—that were previously done manually, leading to a 20% uptick in overall query volume on top of higher per-query costs. While this has temporarily doubled our LLM bill, the time saved on labor-intensive tasks is already reducing overtime costs for senior staff, and we project these efficiency gains will offset the increased LLM expenses and generate net savings by Q4 of this year.
out 741 tok (+581 thinking) latency 11.1 s
Instruction following (exactly three sentences - countable), audience shaping (CFO register), and the hidden-thinking billing gap the token meters below expose.
ByteDance Seed 1.8 passed · 8/8 cases
### Fixed Function ```python def longest_run(xs): if not xs: return 0 best = 1 cur = 1 for i in range(1, len(xs)): if xs[i] == xs[i-1]: cur += 1 best = max(best, cur) else: cur = 1 return best ``` ### Bug Explanation The original function had two critical bugs: 1) It initialized `best` and `cur` to 0, ignoring that any non-empty list has a minimal run length of 1 (causing single-element inputs to return 0 instead of 1). 2) It counted consecutive equal transitions (e.g., for a run of 3 elements, it counted 2 transitions) instead of the actual run length, leading to off-by-one errors for all runs—including those ending at the last index (like [2,2,2] returning 2 instead of 3). […]
out 2072 tok (+1817 thinking) latency 26.1 s
Whether the fix is actually correct (runnable), explanation density, and token efficiency on a bounded task.
ByteDance Seed 1.8 passed · 5/5 fields, guidance "withheld pending the Q4 close"
{ "company": "Meridian Robotics", "quarter": "Q3", "revenue_usd_m": 412, "growth_pct": 11, "guidance": "withheld pending the Q4 close" }
out 757 tok (+695 thinking) latency 10.6 s
Schema obedience (no invented fields), hallucination pressure (guidance is explicitly withheld), and structured-output path differences.
ByteDance Seed 1.8 missed · 147 words, 0 banned, 1 question
Today, we’re launching a cross-provider prompt caching feature for our API gateway, built to solve key challenges for teams leveraging multiple AI service providers. Many teams reuse standardized prompts across tools but incur unnecessary latency and inflated costs from resubmitting identical requests to different platforms. This feature automatically detects and stores matching prompts across all connected providers in an encrypted, user-configurable cache, cutting redundant API calls by up to 55% and reducing average request latency by 40% for repeated prompts. It supports all major AI providers and integrates with existing gateway workflows, with options to set cache expiration times per prompt category. […]
out 1146 tok (+970 thinking) latency 15.2 s
Constraint obedience (word budget, banned-word list, the single question), style fingerprint, and length control.
Use ByteDance Seed 1.8 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="ByteDance-Seed-1.8",
messages=[{"role": "user", "content": "Summarize this diff"}],
)
print(resp.choices[0].message.content)import OpenAI from "openai";
const client = new OpenAI({
baseURL: "https://synthorai.io/v1",
apiKey: "sk-syn-...",
});
const resp = await client.chat.completions.create({
model: "ByteDance-Seed-1.8",
messages: [{ role: "user", content: "Summarize this diff" }],
});
console.log(resp.choices[0].message.content);curl https://synthorai.io/v1/chat/completions \
-H "Authorization: Bearer sk-syn-..." \
-H "Content-Type: application/json" \
-d '{
"model": "ByteDance-Seed-1.8",
"messages": [{"role": "user", "content": "Hello"}]
}'package main
import (
"context"
"fmt"
"github.com/openai/openai-go/v3"
"github.com/openai/openai-go/v3/option"
)
func main() {
client := openai.NewClient(
option.WithBaseURL("https://synthorai.io/v1"),
option.WithAPIKey("sk-syn-..."),
)
resp, _ := client.Chat.Completions.New(context.TODO(), openai.ChatCompletionNewParams{
Model: "ByteDance-Seed-1.8",
Messages: []openai.ChatCompletionMessageParamUnion{
openai.UserMessage("Summarize this diff"),
},
})
fmt.Println(resp.Choices[0].Message.Content)
}import com.openai.client.OpenAIClient;
import com.openai.client.okhttp.OpenAIOkHttpClient;
import com.openai.models.chat.completions.*;
OpenAIClient client = OpenAIOkHttpClient.builder()
.baseUrl("https://synthorai.io/v1")
.apiKey("sk-syn-...")
.build();
ChatCompletion resp = client.chat().completions().create(
ChatCompletionCreateParams.builder()
.model("ByteDance-Seed-1.8")
.addUserMessage("Summarize this diff")
.build());
System.out.println(resp.choices().get(0).message().content().orElse(""));About ByteDance Seed 1.8
- Release notes add upgraded multimodal understanding and more flexible context management.
- It accepts text, image, and video input and returns text, with a 256K context window, 224K maximum input, and 64K output including a 32K chain-of-thought budget, plus function calling, structured output, and context caching.
- Two details are worth setting before production: max output defaults to 4K, well under the 64K ceiling, and thinking.type is enabled by default with no automatic mode, so deep reasoning has to be switched off explicitly if you do not want it.
- Depth is tuned separately through reasoning_effort at minimal, low, medium, or high, with medium the default.
- The generation's headline change is the visual encoder: images up to four megapixels now cost 44.4% of the tokens the previous version charged, high-detail input reaches nine megapixels, and the video frame ceiling doubles to 1,280, though uploading video through the Files API requires naming this model or the older encoder is used.
- Caching works implicitly and explicitly at both prefix and session level, context editing can clear prior thinking or tool calls, and batch inference is supported.
- Synthorai serves it via its OpenAI-compatible gateway endpoint.
FAQ
Is the ByteDance Seed 1.8 API free to try?
Yes: new accounts get 10 trial calls and up to $1 in free credit, no card required. At $0.25/M input tokens, that credit alone covers roughly 500 requests of ~8K tokens against ByteDance Seed 1.8.
What is ByteDance Seed 1.8 best at?
Stronger multimodal understanding and agent capabilities; text, image, and video input; 256K context with 32K chain-of-thought budget. See the About section for the full picture from the vendor's own release notes.
How much does ByteDance Seed 1.8 cost?
ByteDance Seed 1.8 costs $0.25 per million input tokens and $2 per million output tokens on Synthorai. That is the provider's list price, with no platform markup. Cached input tokens bill at $0.05/M.
Does ByteDance Seed 1.8 support prompt caching?
Yes: automatic caching is on by default, with an explicit mode for guaranteed savings. Cached input tokens bill at $0.05/M vs $0.25/M uncached; prompts need a 1,024-token stable prefix to cache. Prompt caching guide →
How do I get access to ByteDance Seed 1.8?
Point your existing OpenAI SDK at base_url="https://synthorai.io/v1", set model="ByteDance-Seed-1.8", 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.