GPT-5.4 Mini

{
  "id": "openai/gpt-5.4-mini",
  "name": "GPT-5.4 mini",
  "description": "Strong small GPT for coding subagents, quick tool use, and high-volume work",
  "family": "gpt-mini",
  "last_updated": "2026-03-17",
  "knowledge": "2025-08-31",
  "attachment": true,
  "reasoning": true,
  "tool_call": true,
  "structured_output": true,
  "temperature": true,
  "modalities": {
    "input": [
      "text",
      "image"
    ],
    "output": [
      "text"
    ]
  },
  "limit": {
    "context": 400000,
    "input": 272000,
    "output": 128000
  },
  "benchmarks": [
    {
      "name": "SWE-Bench Pro",
      "score": 54.4,
      "source": "https://openai.com/index/introducing-gpt-5-4-mini-and-nano/",
      "metric": "resolve rate",
      "variant": "reasoning effort xhigh",
      "date": "2026-03-17"
    },
    {
      "name": "Terminal-Bench",
      "score": 60,
      "source": "https://openai.com/index/introducing-gpt-5-4-mini-and-nano/",
      "metric": "accuracy",
      "variant": "reasoning effort xhigh",
      "version": "2.0",
      "date": "2026-03-17"
    },
    {
      "name": "MCP Atlas",
      "score": 57.7,
      "source": "https://openai.com/index/introducing-gpt-5-4-mini-and-nano/",
      "metric": "score",
      "variant": "reasoning effort xhigh",
      "date": "2026-03-17"
    },
    {
      "name": "Toolathlon",
      "score": 42.9,
      "source": "https://openai.com/index/introducing-gpt-5-4-mini-and-nano/",
      "metric": "score",
      "variant": "reasoning effort xhigh",
      "date": "2026-03-17"
    },
    {
      "name": "τ²-Bench Telecom",
      "score": 93.4,
      "source": "https://openai.com/index/introducing-gpt-5-4-mini-and-nano/",
      "metric": "accuracy",
      "variant": "reasoning effort xhigh",
      "date": "2026-03-17"
    },
    {
      "name": "GPQA Diamond",
      "score": 88,
      "source": "https://openai.com/index/introducing-gpt-5-4-mini-and-nano/",
      "metric": "accuracy",
      "variant": "reasoning effort xhigh",
      "date": "2026-03-17"
    },
    {
      "name": "Humanity's Last Exam",
      "score": 41.5,
      "source": "https://openai.com/index/introducing-gpt-5-4-mini-and-nano/",
      "metric": "accuracy",
      "variant": "with tools",
      "date": "2026-03-17"
    },
    {
      "name": "Humanity's Last Exam",
      "score": 28.2,
      "source": "https://openai.com/index/introducing-gpt-5-4-mini-and-nano/",
      "metric": "accuracy",
      "variant": "without tools",
      "date": "2026-03-17"
    },
    {
      "name": "OSWorld-Verified",
      "score": 72.1,
      "source": "https://openai.com/index/introducing-gpt-5-4-mini-and-nano/",
      "metric": "success rate",
      "variant": "reasoning effort xhigh",
      "date": "2026-03-17"
    },
    {
      "name": "MMMU Pro",
      "score": 78,
      "source": "https://openai.com/index/introducing-gpt-5-4-mini-and-nano/",
      "metric": "accuracy",
      "variant": "with Python",
      "date": "2026-03-17"
    },
    {
      "name": "MMMU Pro",
      "score": 76.6,
      "source": "https://openai.com/index/introducing-gpt-5-4-mini-and-nano/",
      "metric": "accuracy",
      "variant": "without tools",
      "date": "2026-03-17"
    },
    {
      "name": "OmniDocBench",
      "score": 0.1263,
      "source": "https://openai.com/index/introducing-gpt-5-4-mini-and-nano/",
      "metric": "overall edit distance",
      "variant": "reasoning effort none",
      "version": "1.5",
      "date": "2026-03-17"
    },
    {
      "name": "OpenAI MRCR",
      "score": 47.7,
      "source": "https://openai.com/index/introducing-gpt-5-4-mini-and-nano/",
      "metric": "accuracy",
      "variant": "8-needle, 64K-128K",
      "version": "v2",
      "date": "2026-03-17"
    },
    {
      "name": "OpenAI MRCR",
      "score": 33.6,
      "source": "https://openai.com/index/introducing-gpt-5-4-mini-and-nano/",
      "metric": "accuracy",
      "variant": "8-needle, 128K-256K",
      "version": "v2",
      "date": "2026-03-17"
    },
    {
      "name": "Graphwalks",
      "score": 76.3,
      "source": "https://openai.com/index/introducing-gpt-5-4-mini-and-nano/",
      "metric": "accuracy",
      "variant": "BFS, 0-128K",
      "date": "2026-03-17"
    },
    {
      "name": "Graphwalks",
      "score": 71.5,
      "source": "https://openai.com/index/introducing-gpt-5-4-mini-and-nano/",
      "metric": "accuracy",
      "variant": "parents, 0-128K",
      "date": "2026-03-17"
    }
  ],
  "release_date": "2026-03-17",
  "open_weights": false
}