🧠 Reasoning · llm-stats.com
ARC-AGI-2
Second-generation abstraction and reasoning tasks measuring fluid intelligence on novel visual patterns.
9
Modelli testati
GPT-5.5
Modello di punta
85%
Punteggio migliore
OpenAI
Provider
| # | Modello | ARC-AGI-2 |
|---|---|---|
| 1 | GPT-5.5 OpenAI | 85% |
| 2 | Gemini 3.1 Pro Google | 77.1% |
| 3 | GPT-5.4 OpenAI | 73.3% |
| 4 | Claude Opus 4.6 Anthropic | 68.8% |
| 5 | Claude Sonnet 4.6 Anthropic | 58.3% |
| 6 | GPT-5.2 OpenAI | 52.9% |
| 7 | Muse Spark Meta | 42.5% |
| 8 | Gemini 3 Flash Google | 33.6% |
| 9 | Gemini 3 Pro Google | 31.1% |
Dati benchmark aggregati da llm-stats.com. Vedi Metodologia.