Leaderboard
OSWorld
4 models scored · metric: success rate. 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 4 by pts per dollar
- 1.GPT-5.6 Luna203
- 2.GPT-5.6 Terra22.3
- 3.GPT-5.6 Sol11.1
- 4.Claude Opus 57.1
| # | Model | Score | Best in /M | Best out /M | Pts / $ | Context |
|---|---|---|---|---|---|---|
| 🥇 | Claude Opus 5anthropic | 70.6 | $5 | $25 | 7.1 | 1M |
| 🥈 | GPT-5.6 Solopenai | 62.6 | $2.50 | $15 | 11.1 | 1.05M |
| 🥉 | GPT-5.6 Terraopenai | 50.2 | $1 | $6 | 22.3 | 1.05M |
| 4 | GPT-5.6 Lunaopenai | 45.6 | $0.1 | $0.6 | 203 | 1.05M |
Cheapest scorer: GPT-5.6 Luna at $0.1 input /M.