GPT-6 Astra

{
  "id": "openai/gpt-6-astra",
  "name": "GPT-6 Astra",
  "description": "GPT-6 Astra is OpenAI's most capable model for complex reasoning, coding, computer use, research, and document creation.",
  "family": "gpt-astra",
  "last_updated": "2026-09-04",
  "knowledge": "2026-04-30",
  "attachment": true,
  "reasoning": true,
  "tool_call": true,
  "structured_output": true,
  "temperature": false,
  "modalities": {
    "input": [
      "text",
      "image",
      "pdf"
    ],
    "output": [
      "text"
    ]
  },
  "limit": {
    "context": 1050000,
    "input": 922000,
    "output": 128000
  },
  "benchmarks": [
    {
      "name": "Agents' Last Exam",
      "score": 59.3,
      "source": "https://openai.com/index/gpt-6-astra/",
      "date": "2026-09-03"
    },
    {
      "name": "OSWorld",
      "score": 72.6,
      "source": "https://openai.com/index/gpt-6-astra/",
      "metric": "partial score",
      "dataset": "V2-Offline (v2026.08.08)",
      "version": "2.0",
      "date": "2026-09-03"
    },
    {
      "name": "ScreenSpot-Pro",
      "score": 92.7,
      "source": "https://openai.com/index/gpt-6-astra/",
      "metric": "accuracy",
      "variant": "no tools",
      "date": "2026-09-03"
    },
    {
      "name": "AutomationBench",
      "score": 41.4,
      "source": "https://openai.com/index/gpt-6-astra/",
      "metric": "success rate",
      "date": "2026-09-03"
    },
    {
      "name": "BenchCAD",
      "score": 95.9,
      "source": "https://openai.com/index/gpt-6-astra/",
      "metric": "geometric overlap",
      "variant": "with tools",
      "date": "2026-09-03"
    },
    {
      "name": "BrowseComp",
      "score": 91.5,
      "source": "https://openai.com/index/gpt-6-astra/",
      "metric": "accuracy",
      "date": "2026-09-03"
    },
    {
      "name": "Terminal-Bench",
      "score": 57.9,
      "source": "https://openai.com/index/gpt-6-astra/",
      "metric": "success rate",
      "version": "4.0",
      "date": "2026-09-03"
    },
    {
      "name": "DeepSWE",
      "score": 74.1,
      "source": "https://openai.com/index/gpt-6-astra/",
      "metric": "resolve rate",
      "version": "1.1",
      "date": "2026-09-03"
    },
    {
      "name": "FrontierCode",
      "score": 64.5,
      "source": "https://openai.com/index/gpt-6-astra/",
      "metric": "score",
      "dataset": "Extended",
      "version": "1.1",
      "date": "2026-09-03"
    },
    {
      "name": "Terminal-Bench Science",
      "score": 64.6,
      "source": "https://openai.com/index/gpt-6-astra/",
      "metric": "success rate",
      "version": "0.1",
      "date": "2026-09-03"
    },
    {
      "name": "FrontierMath",
      "score": 97.6,
      "source": "https://openai.com/index/gpt-6-astra/",
      "metric": "accuracy",
      "dataset": "Tier 4",
      "version": "v2",
      "date": "2026-09-03"
    },
    {
      "name": "GPQA Diamond",
      "score": 96,
      "source": "https://openai.com/index/gpt-6-astra/",
      "metric": "accuracy",
      "date": "2026-09-03"
    },
    {
      "name": "Humanity's Last Exam",
      "score": 57.2,
      "source": "https://openai.com/index/gpt-6-astra/",
      "metric": "accuracy",
      "variant": "with tools",
      "date": "2026-09-03"
    },
    {
      "name": "ExploitBench",
      "score": 100,
      "source": "https://openai.com/index/gpt-6-astra/",
      "metric": "success rate",
      "variant": "without production safeguards",
      "date": "2026-09-03"
    },
    {
      "name": "SRE-Bench",
      "score": 88,
      "source": "https://openai.com/index/gpt-6-astra/",
      "metric": "success rate",
      "variant": "single attempt",
      "date": "2026-09-03"
    },
    {
      "name": "ARC-AGI-3",
      "score": 99.9,
      "source": "https://openai.com/index/gpt-6-astra/",
      "metric": "RHAE",
      "harness": "Responses API",
      "date": "2026-09-03"
    },
    {
      "name": "Artificial Analysis Intelligence Index",
      "score": 61,
      "source": "https://artificialanalysis.ai/articles/benchmarking-gpt-6-astra",
      "metric": "index score",
      "variant": "max",
      "version": "4.1.1",
      "date": "2026-09-03"
    },
    {
      "name": "Artificial Analysis Coding Agent Index",
      "score": 67,
      "source": "https://artificialanalysis.ai/articles/benchmarking-gpt-6-astra",
      "metric": "index score",
      "harness": "Codex",
      "variant": "max",
      "version": "1.4",
      "date": "2026-09-03"
    }
  ],
  "release_date": "2026-09-04",
  "open_weights": false
}