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

ARC-AGI-2

Independent testSaturatingARC Prize · updated Oct 8, 2026

What it measures

Abstract visual puzzles built to be easy for people and unfamiliar to models, verified on a semi-private set.

How to read it

Now close to saturated at the top, so read it together with cost per task — reaching the same score for a tenth of the price is the more interesting result.

Open at ARC Prize
Best
95.0%
GPT-6 Astra
Median
65.8%
GLM-5.3-Flash

Results

16 of 41 tracked models have a result.

Filter in explorer
  1. 1
    GPT-6 Astra

    OpenAI · Max effort · $1.12 per task

    95.0%
  2. 2
    GPT-6.1 Sol

    OpenAI · Max effort · $0.25 per task

    94.2%
  3. 3
    Claude Opus 5.5

    Anthropic · High effort · $0.41 per task

    93.3%
  4. 4
    Claude Fable 5.1

    Anthropic · Max effort · $4.49 per task

    90.0%
  5. 5
    Gemini 3.8 Flash

    Google DeepMind · High effort · $0.40 per task

    89.2%
  6. 6
    GPT-5.6 Terra

    OpenAI · Max effort · $1.09 per task

    83.9%
  7. 7
    Gemini 3.1 Pro

    Google DeepMind · Preview effort · $0.96 per task

    77.1%
  8. 8
    DeepSeek V4.1 FlashOpen

    DeepSeek · Max effort · $0.13 per task

    72.9%
  9. 9
    GLM-5.3-FlashOpen

    Z.ai · Max effort · $0.09 per task

    65.8%
  10. 10
    Grok 4.7

    SpaceXAI · xHigh effort · $2.01 per task

    61.4%
  11. 11
    DeepSeek V4 ProOpen

    DeepSeek · Max effort · $0.60 per task

    61.3%
  12. 12
    Kimi K3Open

    Moonshot AI · Max effort · $1.59 per task

    60.4%
  13. 13
    GPT-6 Luna

    OpenAI · Max effort · $0.06 per task

    59.3%
  14. 14
    Qwen3.8-27BOpen

    Alibaba · xHigh effort · $0.45 per task

    42.4%
  15. 15
    TM
    Inkling-SmallOpen

    Thinking Machines Lab · xHigh effort · $0.23 per task

    40.1%
  16. 16
    TM
    InklingOpen

    Thinking Machines Lab · $0.64 per task

    36.5%
25 tracked models without a result

Claude Haiku 5.5, Claude Sonnet 5.5, Command A+, Gemini 3.5 Flash-Lite, Gemini 4 Argon, Gemma 4 31B, GLM-5.3, gpt-oss-120b, Hy3, Hy4 preview, Kimi K2.7 Code, 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-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