Benchmarks · reasoning
ARC-AGI-3
ARC-AGI-3 semi-private set: interactive game-like environments where the agent must learn rules on the fly; percent of possible score.
23 models · 69 results · data as of 2026-10-29 · arcprize.org ↗
Leaderboard
GPT-6 Astra high
All results
69 rows| # | Model | Effort | ARC-AGI-3 | Harness |
|---|---|---|---|---|
| 1 | GPT-6 Astra | high | 99.9% | |
| 2 | GPT-6 Astra | max | 98.6% | |
| 3 | GPT-6 Astra | xhigh | 98.4% | |
| 4 | GPT-6 Astra | medium | 98.4% | |
| 5 | GPT-6 Astra | low | 98.0% | |
| 6 | Gemini 4 Argon | low | 97.5% | |
| 7 | Gemini 4 Argon | medium | 97.3% | |
| 8 | Gemini 4 Argon | high | 97.3% | Provider Adapter |
| 9 | GPT-6 Astra | none | 96.7% | |
| 10 | GPT-6.1 Sol | xhigh | 96.4% | |
| 11 | GPT-6.1 Sol | max | 96.2% | |
| 12 | GPT-6.1 Sol | high | 95.0% | |
| 13 | GPT-6.1 Sol | medium | 91.0% | |
| 14 | GPT-6.1 Sol | low | 82.8% | |
| 15 | Sonnet 5.5 | xhigh | 72.4% | |
| 16 | Sonnet 5.5 | high | 35.5% | |
| 17 | Gemini 3.8 Flash | high | 35.0% | |
| 18 | Opus 5 | high | 30.2% | |
| 19 | Gemini 3.8 Flash | medium | 24.2% | |
| 20 | GPT-6 Sol | max | 23.0% | |
| 21 | Gemini 3.8 Flash | low | 15.1% | |
| 22 | Grok 4.7 | high | 10.0% | |
| 23 | GPT-6 Sol | xhigh | 9.5% | |
| 24 | Grok 4.7 | medium | 8.4% | |
| 25 | GPT-5.6 Sol | max | 7.8% | |
| 26 | Grok 4.7 | xhigh | 7.0% | |
| 27 | GPT-5.6 Sol | xhigh | 7.0% | |
| 28 | GPT-6 Sol | high | 5.9% | |
| 29 | GPT-6 Sol | medium | 4.9% | |
| 30 | Sonnet 5.5 | medium | 4.9% |