NezhaGateNezhaGate
Chat

Doubao Seed 2.1 Turbo

⧉
doubao-seed-2-1-turbo
Model Type ProTurbo

Doubao Seed 2.1 Turbo is the lower-cost member of the Doubao Seed 2.1 family, at about half the unit price of Pro, for everyday chat and high-volume jobs. By default it thinks before it answers, with the reasoning in reasoning_content billed at the output rate (reasoning_effort: "minimal" usually skips it, though less reliably than on Pro); tool calling, JSON output, streaming and image input (base64 or public links) are supported. A repeated long prefix can hit the cache automatically, and that part settles at the cache rate. Fully OpenAI-compatible: set model to doubao-seed-2-1-turbo. Pay-as-you-go, failed calls never billed.

ChatReasoningLow costPrompt cachingOpenAI compatible

Live Test · Playground

Try out Doubao Seed 2.1 Turbo right here (available after login).

Input

Advanced
Web search
Memory · multi-turn

Conversation

Start a conversationType a message below to begin

About Doubao Seed 2.1 Turbo

Doubao Seed 2.1 Turbo is the lower-cost member of the Doubao Seed 2.1 family, served on NezhaGate through the OpenAI-compatible API. It costs about half as much per token as Pro, which suits everyday chat and high-volume jobs, and it accepts image input. By default it thinks before it answers: the reasoning comes back in message.reasoning_content and the answer in content, and reasoning tokens bill at the output rate. Tool calling, JSON output, streaming and prompt caching are supported. Set model to doubao-seed-2-1-turbo; pay-as-you-go, failed calls never billed.

Use cases

Everyday chat and Q&A

A low unit price for high-frequency chat such as support, Q&A and everyday writing; switching models is just the model field.

High-volume text processing

Classification, extraction, summarising and rewriting in bulk, at about half the unit price of Pro.

Image understanding

Q&A over screenshots, receipts and charts; in our tests it read the text in an image accurately. Send images as base64 or public links.

Repeated questions on one long document

A repeated long prefix hits the cache automatically (9,528 of 9,828 input tokens on the second call in our tests), and the cached part settles at the cache rate.

How to choose

Choose Turbo for everyday chat, high-volume jobs or when the unit price matters most, with automatic caching of a repeated long prefix; choose Pro for the flagship reasoning, coding and agent work of the Doubao Seed 2.1 series. Both belong to the Doubao Seed 2.1 family, so switching is just the model field.

FAQ

Can I turn thinking off?
Usually, though less reliably than on Pro. With reasoning_effort: "minimal", 8 of 12 calls in our tests skipped the thinking and answered directly, while the other 4 still returned reasoning; thinking: {"type": "disabled"} has no effect on this model. When omitted, the model thinks first, and reasoning tokens bill at the output rate inside usage.completion_tokens.
Can max_tokens limit the output?
No. max_tokens is accepted but does not cut the output off (with max_tokens 60 we still got 5,929 tokens), and you are billed for the actual tokens in usage. For short answers, ask for brevity in the prompt.
Is prompt caching supported?
Yes. A repeated long prefix hits the cache automatically: in our tests 9,528 of the 9,828 input tokens of the second call were cached, and that part settles at the cache rate. Whether it hits is decided upstream, and billing follows the returned usage.
How do I choose between Turbo and Pro?
Turbo costs about half as much per token, but it sometimes thinks longer on simple prompts: the same five-word greeting used 412-3,656 tokens on Turbo and 362-802 on Pro in our tests, and what you pay follows usage. Turbo saves most on high-volume jobs with long inputs and short outputs (where automatic caching also helps); choose Pro for flagship reasoning, coding and agent work.
Am I charged for failed calls?
No. Only calls that return normally are settled, at actual usage; upstream errors and timeouts never touch your balance.

Related Models

Explore other models you can integrate.

View all →