Gemini 3.8 Flash

{
  "id": "google/gemini-3.8-flash",
  "name": "Gemini 3.8 Flash",
  "description": "Google's most intelligent Flash model, engineered for long-horizon software engineering, autonomous agents, and complex enterprise workflows",
  "family": "gemini-flash",
  "last_updated": "2026-09-02",
  "attachment": true,
  "reasoning": true,
  "tool_call": true,
  "structured_output": true,
  "temperature": true,
  "modalities": {
    "input": [
      "text",
      "image",
      "video",
      "audio",
      "pdf"
    ],
    "output": [
      "text"
    ]
  },
  "limit": {
    "context": 1048576,
    "output": 65536
  },
  "benchmarks": [
    {
      "name": "DeepSWE",
      "score": 73.7,
      "source": "https://blog.google/innovation-and-ai/models-and-research/gemini-models/3-8-flash-and-3-8-flash-cyber/",
      "metric": "pass@1",
      "harness": "mini-swe-agent",
      "variant": "high thinking",
      "version": "1.1"
    },
    {
      "name": "GDPval-AA",
      "score": 1545,
      "source": "https://blog.google/innovation-and-ai/models-and-research/gemini-models/3-8-flash-and-3-8-flash-cyber/",
      "metric": "Elo",
      "version": "2"
    },
    {
      "name": "Finance Agent",
      "score": 61.4,
      "source": "https://blog.google/innovation-and-ai/models-and-research/gemini-models/3-8-flash-and-3-8-flash-cyber/",
      "metric": "pass@1",
      "version": "2"
    },
    {
      "name": "Harvey Legal Agent Benchmark",
      "score": 10,
      "source": "https://blog.google/innovation-and-ai/models-and-research/gemini-models/3-8-flash-and-3-8-flash-cyber/",
      "metric": "all-pass rate"
    },
    {
      "name": "Terminal-Bench",
      "score": 89.4,
      "source": "https://blog.google/innovation-and-ai/models-and-research/gemini-models/3-8-flash-and-3-8-flash-cyber/",
      "metric": "pass@1",
      "harness": "Terminus 2",
      "version": "2.1"
    },
    {
      "name": "Terminal-Bench",
      "score": 19.1,
      "source": "https://blog.google/innovation-and-ai/models-and-research/gemini-models/3-8-flash-and-3-8-flash-cyber/",
      "metric": "pass@1",
      "version": "4.0"
    },
    {
      "name": "GDP.PDF",
      "score": 35,
      "source": "https://blog.google/innovation-and-ai/models-and-research/gemini-models/3-8-flash-and-3-8-flash-cyber/",
      "metric": "all-pass rate"
    },
    {
      "name": "CharXiv Reasoning",
      "score": 86.2,
      "source": "https://blog.google/innovation-and-ai/models-and-research/gemini-models/3-8-flash-and-3-8-flash-cyber/",
      "metric": "pass@1",
      "variant": "without tools"
    },
    {
      "name": "LVBench",
      "score": 87.8,
      "source": "https://blog.google/innovation-and-ai/models-and-research/gemini-models/3-8-flash-and-3-8-flash-cyber/",
      "metric": "pass@1",
      "variant": "agentic"
    },
    {
      "name": "LVBench",
      "score": 87.1,
      "source": "https://blog.google/innovation-and-ai/models-and-research/gemini-models/3-8-flash-and-3-8-flash-cyber/",
      "metric": "pass@1",
      "variant": "static; 1024 frames; without tools"
    },
    {
      "name": "HLE-Verified",
      "score": 54.9,
      "source": "https://blog.google/innovation-and-ai/models-and-research/gemini-models/3-8-flash-and-3-8-flash-cyber/",
      "metric": "accuracy",
      "dataset": "1811 verified and revised items"
    },
    {
      "name": "OSWorld",
      "score": 59,
      "source": "https://blog.google/innovation-and-ai/models-and-research/gemini-models/3-8-flash-and-3-8-flash-cyber/",
      "metric": "partial score",
      "harness": "Gemini CUA",
      "variant": "batched tool calls; best of 3 runs; 500 steps",
      "dataset": "before 2026-08-08 patch",
      "version": "2.0"
    },
    {
      "name": "BioMysteryBench",
      "score": 88.8,
      "source": "https://blog.google/innovation-and-ai/models-and-research/gemini-models/3-8-flash-and-3-8-flash-cyber/",
      "metric": "pass@1",
      "variant": "with tools",
      "dataset": "human-solvable"
    },
    {
      "name": "BioMysteryBench",
      "score": 56.5,
      "source": "https://blog.google/innovation-and-ai/models-and-research/gemini-models/3-8-flash-and-3-8-flash-cyber/",
      "metric": "pass@1",
      "variant": "with tools",
      "dataset": "human-difficult"
    },
    {
      "name": "LAB-Bench",
      "score": 86.2,
      "source": "https://blog.google/innovation-and-ai/models-and-research/gemini-models/3-8-flash-and-3-8-flash-cyber/",
      "metric": "macro-average",
      "variant": "with tools",
      "dataset": "11 subtasks",
      "version": "2"
    }
  ],
  "release_date": "2026-09-02",
  "open_weights": false
}