# Prompts for Doubao: goal and scenario, not step-by-step

> How to prompt Doubao Seed 2.1 (Pro, Lite, Turbo) on ByteDance's Volcengine Ark: give the goal and scenario, avoid step-by-step decomposition, put instructions in the user message and raise max_tokens.

Source: Promptea — https://www.promptea.me/en/models/doubao · Language: English · Last updated: 2026-11-14

## Quick tips

- Volcengine's guide for reasoning models says to add the goal and the scenario (who it is for, what it is used for, what a good outcome is) to the problem itself.
- It also says to reduce or avoid breaking the problem into steps or giving examples: with thinking on, that limits the model's own reasoning.
- Use fewer system prompts: send the prompt information as the user message.
- Default max output is 4K tokens: raise max_tokens for long answers. For JSON (beta), json_object needs the word 'json' in the input; json_schema with strict enforces a schema.

## Current models Promptea supports

Capabilities come from the provider’s official documentation; each entry shows the date it was last verified.

### Doubao Seed 2.1 Pro (default)

The most capable Doubao on Volcengine Ark (Sep 2026), aimed at coding and agents: 1M (1024K) context, multimodal input and thinking on by default. On BytePlus (international) the family is called Dola Seed.

- Context window: 1.05M tokens · max output: 262K
- Input: text, image, video · output: text
- Latency: slower (deliberate) · structured output: yes · tool use: yes · native search: yes
- Strengths: coding, reasoning, long context, multimodal
- How to prompt it: Reasons by default (thinking type disabled turns it off; reasoning_effort defaults to high). Default max output is 4K: raise max_tokens for long answers. For JSON, json_schema with strict, or json_object with the word 'json' in the text.
- Verified: 2026-11-14 — [source](https://docs.volcengine.com/docs/82379/1330310)

### Doubao Seed 2.1 Lite

Lower-cost Doubao (Sep 2026) with the same 1M context; it also accepts audio.

- Context window: 1.05M tokens · max output: 262K
- Input: text, image, video, audio · output: text
- Latency: balanced · structured output: yes · tool use: yes · native search: no
- Strengths: speed, multimodal, long context
- How to prompt it: Reasons by default; turn it off (thinking type disabled) for simple tasks. Raise max_tokens if you expect a long answer (the default is 4K).
- Verified: 2026-11-14 — [source](https://docs.volcengine.com/docs/82379/1330310)

### Doubao Seed 2.1 Turbo

Low-latency Doubao (Jun 2026): 256K context. For fast answers and high volume.

- Context window: 262K tokens · max output: 262K
- Input: text, image, video · output: text
- Latency: fast · structured output: yes · tool use: yes · native search: no
- Strengths: speed
- How to prompt it: For short tasks, turn reasoning off (thinking type disabled) and ask for a short, exact output format.
- Verified: 2026-11-14 — [source](https://docs.volcengine.com/docs/82379/1330310)

## Templates

### Goal + scenario request

```text
Problem: [what you need]
Goal: [what a good result lets you do]
Scenario: [who will use it, where and when]
Audience: [who reads it and what they already know]
Constraints: [length, tone, must-nots]
Output format: [exact shape]
```

[Open in Promptea](https://www.promptea.me/en/analyzer?prompt=Problem%3A+%5Bwhat+you+need%5D%0AGoal%3A+%5Bwhat+a+good+result+lets+you+do%5D%0AScenario%3A+%5Bwho+will+use+it%2C+where+and+when%5D%0AAudience%3A+%5Bwho+reads+it+and+what+they+already+know%5D%0AConstraints%3A+%5Blength%2C+tone%2C+must-nots%5D%0AOutput+format%3A+%5Bexact+shape%5D&purpose=text&target=doubao)

## Frequently asked questions

### Doubao or Dola Seed?

The same family under two brands: Doubao on Volcengine Ark (China), Dola Seed on BytePlus ModelArk (international). Promptea lists the Volcengine model ids.

### Why is my Doubao answer cut short?

The default max output is 4K tokens even though the models allow up to 256K. Set max_tokens explicitly for long answers.
