Sci-MMR (Agentic Tool-Use): leaderboard
The same Sci-MMR tasks solved as a tool-using agent (crop, zoom, search over the figures); accuracy averaged over difficulty tiers. Higher is better. Source: arxiv.org/abs/2609.11243.
Metric: Accuracy (%). Source: arxiv.org. Saturation forecast: Around February 2027. 8 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | GPT-5.5 | 69.8 | #33 |
| 2 | Claude Opus 4.8 | 60 | #39 |
| 3 | Kimi K2.7 Code | 52.8 | #95 |
| 4 | Gemini 3.1 Pro (Preview) | 52.3 | #64 |
| 5 | MiniMax-M3 | 50.6 | #151 |
| 6 | GLM-5V Turbo | 44.3 | #172 |
| 7 | Qwen 3.7 Plus | 40.9 | #98 |
No result here: #3 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=sci-mmr-agentic-tool-use · How It Works · Data refreshed daily, snapshot 2026-11-05.