How do I switch from OpenAI to DeepSeek -- is it really a drop-in replacement?

Yes. DeepSeek V4 supports the OpenAI chat completions format. Point OpenClaw at RunAPI's /v1 base URL, change the model parameter to deepseek-v4-pro or deepseek-v4-flash, and the rest of your config stays the same. No code changes beyond the model name.

What are thinking tokens in DeepSeek and how do they affect cost?

When you enable reasoning (thinking.type set to enabled), DeepSeek generates additional "thinking" tokens that show its chain-of-thought process. These tokens count as output tokens and are billed accordingly. They improve accuracy on complex tasks but increase total token usage.

What is the difference between deepseek-v4-flash and deepseek-v4-pro?

Flash prioritizes speed with lower latency and cost. Pro prioritizes quality with better performance on complex reasoning, coding, and multi-step tasks. Both use the same endpoint and accept the same parameters.

Does RunAPI route DeepSeek requests through China?

No. RunAPI proxies DeepSeek through its own infrastructure. Your requests do not route through Chinese data centers, and your API key is never shared with DeepSeek directly.

How does DeepSeek pricing on RunAPI compare to direct DeepSeek API pricing?

RunAPI prices DeepSeek tokens at 50% of official DeepSeek rates. DeepSeek is already one of the lowest-cost LLM providers, so through RunAPI you get the lowest per-token price available for a reasoning-capable model. Billing is per-token with no subscription or minimum spend.

OPENCLAW + DEEPSEEK

在 OpenClaw 中使用 DeepSeek。

DeepSeek V4 通过 RunAPI 同时提供 OpenAI 兼容和 Anthropic 兼容协议。Flash 主打速度，Pro 主打质量。内置推理模式通过一个请求参数激活——无需单独的模型 slug。RunAPI 以官方 DeepSeek 费率的 50% 对 DeepSeek token 定价。

获取 API Key 阅读文档

一个 API 密钥 · OpenAI + Anthropic 协议 · 支持流式传输

使用 RunAPI 通过 OpenAI 兼容的聊天补全端点调用 DeepSeek V4 Pro。

要求：
- 调用 RunAPI 聊天补全端点 https://runapi.ai/v1/chat/completions
- 将 model 设置为 "deepseek-v4-pro"
- 使用 RUNAPI_API_KEY 环境变量进行授权
- 响应是同步的——根据 stream 参数，回复以 SSE 事件流式返回，或作为单个 JSON 对象返回
- 如需启用推理，在请求体中将 thinking.type 设置为 "enabled"

curl -X POST https://runapi.ai/v1/chat/completions \
  -H "Authorization: Bearer $RUNAPI_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "deepseek-v4-pro",
    "messages": [
      {"role": "system", "content": "You are a helpful coding assistant."},
      {"role": "user", "content": "Write a Python function that finds the longest palindromic substring in O(n) time using Manacher algorithm."}
    ],
    "stream": true,
    "max_tokens": 4096
  }'

{
  "id": "chatcmpl-abc123",
  "object": "chat.completion",
  "model": "deepseek-v4-pro",
  "choices": [
    {
      "index": 0,
      "message": {
        "role": "assistant",
        "content": "Here is Manacher's algorithm implementation..."
      },
      "finish_reason": "stop"
    }
  ],
  "usage": {
    "prompt_tokens": 42,
    "completion_tokens": 356,
    "total_tokens": 398
  }
}

复制 curl 命令进行测试 deepseek

工作原理

三步在 OpenClaw 中使用 DeepSeek

Configure RunAPI as provider

Set the RUNAPI_API_KEY environment variable. In OpenClaw, add RunAPI as an OpenAI-compatible provider with baseUrl https://runapi.ai/v1. DeepSeek works through the same provider you use for GPT or Claude — no separate provider entry needed.

export RUNAPI_API_KEY=runapi_xxx

Select a DeepSeek model

Set the model to deepseek-v4-pro for quality or deepseek-v4-flash for speed. Both use the /v1/chat/completions endpoint. To enable reasoning, pass thinking.type set to enabled — no model slug change required.

model: deepseek-v4-pro

Stream the response

OpenClaw receives the response as an SSE stream in OpenAI chat completions format. Token usage is reported in the final event for billing transparency. The same format works whether you use DeepSeek, GPT, or Claude through RunAPI.

stream: true

参数

DeepSeek 聊天补全参数

参数	类型	说明
`model`	`string`	Required. deepseek-v4-pro or deepseek-v4-flash.
`messages`	`array`	Required. Array of message objects with role (system, user, assistant) and content fields.
`stream`	`boolean`	Optional. Set to true for SSE streaming. Default false.
`max_tokens`	`integer`	Optional. Maximum number of tokens to generate.
`temperature`	`number`	Optional. Sampling temperature between 0 and 2. Lower values are more deterministic.
`thinking`	`object`	Optional. Set type to "enabled" to activate DeepSeek's built-in reasoning mode.

OpenClaw 上的 DeepSeek 是什么？

DeepSeek V4 是 OpenAI 兼容的 LLM，在 OpenClaw 中可作为 GPT 的直接替代品——只需修改模型名称，其余配置不变。它通过请求参数支持内置的思维链推理，RunAPI 以 DeepSeek 已经很低的官方价格的 50% 提供，成为推理能力模型中单 token 成本最低的选择。