Omni2Web - Instruction Recovery Action F1: leaderboard
Metric: Action F1 (%; Track B component: recovered edit actions matched one-to-one to the reference actions; 918 bilingual screen recordings of weakly referential webpage edit requests; the model rewrites the recording into an ordered list of explicit target and action instructions without the source HTML; one-to-one matching against reference instructions, semantic equivalence judged by Gemini 3.1 Pro; unusable outputs score zero). Source: arxiv.org. Saturation forecast: Around 2031. 17 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | Claude Opus 5 | 49.76 |
| 2 | GPT-5.6 Sol | 49.17 |
| 3 | Qwen 3.5 Omni Plus | 48.29 |
| 4 | Gemini 3.5 Flash | 47.88 |
| 5 | Grok 4.6 | 47.06 |
| 6 | Kimi K3 | 45.25 |
| 7 | Seed 2.0 Lite | 45.14 |
| 8 | Seed 2.1 Pro | 44.96 |
| 9 | Gemini 3.1 Pro (Preview) | 44.18 |
| 10 | Qwen 3.8 Max | 41.9 |
| 11 | MiMo-V2.5 | 40.67 |
| 12 | Qwen 3.8 27B | 39.85 |
| 13 | MiniMax-M3 | 37.38 |
| 14 | Muse Spark 1.1 | 32.37 |
| 15 | MiniCPM-o-4.5 | 31.93 |
Interactive version: theaggregate.ai/benchmark?slug=omni2web-instruction-recovery-action-f1 · How It Works · Data refreshed daily, snapshot 2026-09-26.