Benchmark
ARC-AGI v2
ARC-AGI-2 is an upgraded benchmark for measuring abstract reasoning and problem-solving abilities in AI systems through visual grid transformation tasks…
- Modality
- multimodal
- Categories
- reasoning, spatial_reasoning, vision
- Openness
- unknown
- Source
- llm_stats
- Reported scores
- 17
Reported scores
llm_stats
| Model | Organization | Reported value | Reported | Evidence |
|---|---|---|---|---|
| Claude Opus 4 | Anthropic | 0.086 | 2025-05-22 | evidence |
| Claude Opus 4.5 | Anthropic | 0.376 | 2025-11-24 | evidence |
| Claude Opus 4.6 | Anthropic | 0.688 | 2026-02-05 | evidence |
| Claude Sonnet 4.6 | Anthropic | 0.583 | 2026-02-17 | evidence |
| Gemini 2.5 Pro | 0.049 | 2025-05-20 | evidence | |
| Gemini 3 Flash | 0.336 | 2025-12-17 | evidence | |
| Gemini 3 Pro | 0.311 | 2025-11-18 | evidence | |
| Gemini 3.1 Pro | 0.771 | 2026-02-19 | evidence | |
| Gemini 3.5 Flash | 0.721 | 2026-05-19 | evidence | |
| GPT-5.2 | OpenAI | 0.529 | 2025-12-11 | evidence |
| GPT-5.2 Pro | OpenAI | 0.542 | 2025-12-11 | evidence |
| GPT-5.4 | OpenAI | 0.733 | 2026-03-05 | evidence |
| GPT-5.5 | OpenAI | 0.85 | 2026-04-23 | evidence |
| Grok-4 | xAI | 0.159 | 2025-07-09 | evidence |
| Inkling-Small | Thinking Machines Lab | 0.401 | 2026-07-30 | evidence |
| Muse Spark | Meta | 0.425 | 2026-04-08 | evidence |
| o3 | OpenAI | 0.065 | 2025-04-16 | evidence |
Scores are partitioned by the source that reported them and are never merged into a single cross-source ranking, because the sources measure different things and say so.