Skip to content
Model Index
  • Overview
  • Models
  • Compare
  • Benchmarks
  • Finder
LabsMethodology
  1. Benchmarks
  2. Reasoning & science

ARC-AGI-3

Independent testARC Prize · updated Oct 8, 2026

What it measures

Interactive, game-like environments with no instructions: the agent has to explore, infer the goal and learn the rules as it plays. 100% means beating every game as efficiently as a person.

How to read it

The newest and least covered benchmark here — only a few models have verified runs. Figures are raw model results; runs using a provider-built adapter harness score far higher and are excluded.

Open at ARC Prize
Best
62.7%
GPT-6 Astra
Median
1.8%
Grok 4.7

Results

6 of 41 tracked models have a result.

Filter in explorer
  1. 1
    GPT-6 Astra

    OpenAI · Max effort

    62.7%
  2. 2
    GPT-6.1 Sol

    OpenAI · Max effort

    52.7%
  3. 3
    Gemini 3.8 Flash

    Google DeepMind · High effort

    10.4%
  4. 4
    Grok 4.7

    SpaceXAI · xHigh effort

    1.8%
  5. 5
    GPT-5.6 Terra

    OpenAI · Max effort

    0.8%
  6. 6
    GPT-6 Luna

    OpenAI · Medium effort

    0.2%
35 tracked models without a result

Claude Fable 5.1, Claude Haiku 5.5, Claude Opus 5.5, Claude Sonnet 5.5, Command A+, DeepSeek V4 Pro, DeepSeek V4.1 Flash, Gemini 3.1 Pro, Gemini 3.5 Flash-Lite, Gemini 4 Argon, Gemma 4 31B, GLM-5.3, GLM-5.3-Flash, gpt-oss-120b, Hy3, Hy4 preview, Inkling, Inkling-Small, Kimi K2.7 Code, Kimi K3, MAI-Thinking-1, MiMo-V2.6-Flash, MiMo-V2.6-Pro, MiniMax-M3, Mistral Large 4, Mistral Medium 3.5, Mistral Small 4, Muse Glimmer 30B, Muse Spark 1.3, Nemotron 3 Ultra, Nova 2 Lite, Qwen3.8-27B, Qwen3.8-Flash, Qwen3.8-Max, Step 5 Preview.

Missing means the evaluator has not published a run, not that the model would score poorly.

Model Index

An independent snapshot of frontier SI models and how they perform on third-party benchmarks.

Data as of Oct 8, 2026

Explore

  • All models
  • Compare
  • Benchmarks
  • Model finder
  • Labs
  • Methodology & sources

Sources

Benchmark results come from Artificial Analysis, Arena, Vals AI and ARC Prize. Specifications and prices come from each lab's own documentation.

Not affiliated with any lab or evaluator. Names and logos are trademarks of their respective owners and are used only to identify them. Scores change often — check the linked source before making a decision that depends on one.

  • Overview
  • Models
  • Compare
  • Benchmarks
  • Finder
  • Overview
  • Models
  • Compare
  • Benchmarks
  • Finder
  • Overview
  • Models
  • Compare
  • Benchmarks
  • Finder