promptfoo--promptfoo/site/docs/providers/deepseek.md

sidebar_label, description

sidebar_label	description
DeepSeek	Configure DeepSeek's OpenAI-compatible API with V4 chat and reasoning models, 1M context windows, and prompt caching for cost-effective LLM testing

DeepSeek

DeepSeek provides an OpenAI-compatible API for their language models, with specialized models for both general chat and advanced reasoning tasks. The DeepSeek provider is compatible with all the options provided by the OpenAI provider.

Setup

1. Get an API key from the DeepSeek Platform
2. Set DEEPSEEK_API_KEY environment variable or specify apiKey in your config

Configuration

Basic configuration example:

providers:
  - id: deepseek:deepseek-chat
    config:
      temperature: 0.7
      max_tokens: 4000
      apiKey: YOUR_DEEPSEEK_API_KEY

  - id: deepseek:deepseek-reasoner # Legacy alias for V4 Flash thinking mode
    config:
      max_tokens: 8000

Configuration Options

• temperature
• max_tokens
• cost, inputCost, outputCost - Override promptfoo's pricing estimates (inputCost and outputCost take precedence over cost)
• top_p, presence_penalty, frequency_penalty
• stream
• showThinking - Control whether reasoning content is included in the output (default: true, applies to deepseek-reasoner model)

Available Models

:::note

The current primary API model names are deepseek-v4-flash and deepseek-v4-pro. The legacy aliases deepseek-chat and deepseek-reasoner remain available until July 24, 2026 and currently map to the non-thinking and thinking modes of deepseek-v4-flash, respectively.

:::

deepseek-v4-flash

• General purpose V4 model for conversations and reasoning
• 1M context window, up to 384K output tokens
• Input: $0.0028/1M (cache hit), $0.14/1M (cache miss)
• Output: $0.28/1M

deepseek-v4-pro

• Higher-capability V4 model with thinking and non-thinking modes
• 1M context window, up to 384K output tokens
• Input: $0.003625/1M (cache hit), $0.435/1M (cache miss)
• Output: $0.87/1M
• Promotional pricing is documented through May 31, 2026

Legacy aliases

deepseek-chat

• Legacy alias that currently maps to non-thinking deepseek-v4-flash
• Scheduled for retirement on July 24, 2026

deepseek-reasoner

• Legacy alias that currently maps to thinking deepseek-v4-flash
• Scheduled for retirement on July 24, 2026
• Supports showing or hiding reasoning content through the showThinking parameter

:::warning

Thinking mode does not support temperature, top_p, presence_penalty, frequency_penalty, logprobs, or top_logprobs parameters. Setting these parameters will not trigger an error but will have no effect.

:::

Example Usage

Here's an example comparing DeepSeek with OpenAI on reasoning tasks:

providers:
  - id: deepseek:deepseek-reasoner
    config:
      max_tokens: 8000
      showThinking: true # Include reasoning content in output (default)
  - id: openai:o-1
    config:
      temperature: 0.0

prompts:
  - 'Solve this step by step: {{math_problem}}'

tests:
  - vars:
      math_problem: 'What is the derivative of x^3 + 2x with respect to x?'

Controlling Reasoning Output

The legacy deepseek-reasoner alias uses V4 Flash thinking mode and includes detailed reasoning steps in its output. You can control whether this reasoning content is shown using the showThinking parameter:

providers:
  - id: deepseek:deepseek-reasoner
    config:
      showThinking: false # Hide reasoning content from output

When showThinking is set to true (default), the output includes both reasoning and the final answer in a standardized format:

Thinking: <reasoning content>

<final answer>

When set to false, only the final answer is included in the output. This is useful when you want better reasoning quality but don't want to expose the reasoning process to end users or in your assertions.

See our complete example that benchmarks it against OpenAI's o1 model on the MMLU reasoning tasks.

API Details

• Base URL: https://api.deepseek.com/v1
• OpenAI-compatible API format
• Full API documentation

See Also

• OpenAI Provider - Compatible configuration options
• Complete example - Benchmark against OpenAI's o1 model