PointerBench - Sheets: leaderboard
PointerBench: 1,500 screenshot/instruction pairs, 500 per suite, with 1024×768 absolute pixel coordinates. One JSON prompt; success requires a point inside the target or box coverage ≥0.90 and precision ≥0.70. Disabled uses 512 output tokens; Low and Medium use 2,048, with a 1,024-token reasoning allowance for Medium. Outputs are unmodified; malformed, filtered and exhausted responses fail without retry. Explicit thinking settings are retained. Success rate (%).
Metric: Success rate (%). Source: arxiv.org. Saturation forecast: Around January 2027. 45 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Claude Opus 5 | 98.8 | #13 |
| 2 | GPT-5.6 Sol (Medium) | 97.8 | #19 (GPT-5.6 Sol) |
| 3 | Claude Fable 5 (Medium) | 97.6 | #17 (Claude Fable 5) |
| 4 | GPT-5.6 Sol (Low) | 96.4 | #19 (GPT-5.6 Sol) |
| 5 | Claude Fable 5 (Low) | 95.8 | #17 (Claude Fable 5) |
| 6 | GPT-5.6 Luna (Medium) | 92 | #75 (GPT-5.6 Luna) |
| 7 | GPT-5.6 Pro Sol | 92 | #20 |
| 8 | GPT-5.6 Terra | 91.6 | #40 |
| 9 | Claude Sonnet 4.6 (Medium) | 88.8 | #94 (Claude Sonnet 4.6) |
| 10 | GPT-5.6 Luna (Low) | 87.6 | #75 (GPT-5.6 Luna) |
| 11 | GPT-5.6 Sol (Non-reasoning) | 85.6 | #19 (GPT-5.6 Sol) |
| 12 | Claude Sonnet 4.6 (Non-reasoning) | 83 | #94 (Claude Sonnet 4.6) |
| 13 | Claude Opus 4.8 | 82.2 | #37 |
| 14 | Claude Sonnet 4.6 (Low) | 79.8 | #94 (Claude Sonnet 4.6) |
| 15 | Claude Sonnet 5 | 78.6 | #62 |
No result here: #5 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=pointerbench-sheets · How It Works · Data refreshed daily, snapshot 2026-10-25.