Leaderboard
ARC-AGI-2
6 models scored · metric: accuracy. Prices are the cheapest listed across all serving providers. Pts/$ = score ÷ blended price (3×input + output) / 4 — how much score each dollar buys.
Best value · top 5 by pts per dollar
- 1.Gemini 3.5 Flash21.4
- 2.Gemini 3.1 Pro Preview17.1
- 3.GPT-5.414.2
- 4.Claude Opus 59.0
- 5.GPT-5.58.3
| # | Model | Score | Best in /M | Best out /M | Pts / $ | Context |
|---|---|---|---|---|---|---|
| 🥇 | Claude Opus 5anthropic | 90.4 | $5 | $25 | 9.0 | 1M |
| 🥈 | GPT-5.5openai | 85 | $4.55 | $27.27 | 8.3 | 1.05M |
| 🥉 | GPT-5.4 Proopenai | 83.3 | $27 | $160 | 1.4 | 1.05M |
| 4 | Gemini 3.1 Pro Previewgoogle | 77.1 | $2 | $12 | 17.1 | 1.05M |
| 5 | GPT-5.4openai | 73.3 | $2.20 | $14 | 14.2 | 1.05M |
| 6 | Gemini 3.5 Flashgoogle | 72.1 | $1.50 | $9 | 21.4 | 1.05M |
Cheapest scorer: Gemini 3.5 Flash at $1.50 input /M.