DeepSeek V4 Flash 0731 on DeepInfra

Date: 1 Aug 2026 · OpenClaw 2026.7.1-2

We added DeepInfra as a native provider in OpenClaw and hooked up DeepSeek V4 Flash (build 0731) behind the alias v4flash. DeepInfra exposes an OpenAI-compatible endpoint, so this was config, not code:

{
  "providers": {
    "deepinfra": {
      "baseUrl": "https://api.deepinfra.com/v1/openai",
      "api": "openai-completions",
      "apiKeyEnv": "DEEPINFRA_API_KEY",
      "models": {
        "deepseek-ai/DeepSeek-V4-Flash-0731": {
          "name": "DeepSeek V4 Flash (0731)",
          "reasoning": true,
          "contextWindow": 1000000,
          "pricePerMTokens": { "input": 0.09, "output": 0.18 }
        }
      }
    }
  },
  "aliases": {
    "v4flash": "deepinfra/deepseek-ai/DeepSeek-V4-Flash-0731"
  }
}

Setup was smooth and so far no hiccups. Pin it to any session with /model v4flash.