OmniStepBench - Multimodal Multi-Step: leaderboard
Audited task accuracy in a shared OpenClaw harness over 233 tasks selected from 22 benchmarks. Six modality/execution groups and task-weighted text/multimodal means remain separate. API and local deployment provenance is retained; text-only models have no multimodal score.
Metric: Audited Accuracy (%). Source: seerray-lab.github.io. Source: arxiv.org. Saturation forecast: Around January 2027. 34 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | GPT-6 Astra | 94.2 | #5 |
| 2 | Claude Opus 5.5 | 92.3 | #3 |
| 3 | GPT-6.1 Sol | 92.3 | #8 |
| 4 | Gemini 3.1 Pro (Preview) | 88.5 | #56 |
| 5 | Claude Opus 4.6 | 76.2 | #60 |
| 6 | GPT-5.5 | 75 | #32 |
| 7 | GPT-5.4 | 71.2 | #78 |
| 8 | Claude Opus 4.8 | 69.2 | #38 |
| 9 | Kimi K3 | 69.2 | #30 |
| 10 | Qwen 3.5 Plus | 69.2 | #124 |
| 11 | MiMo-V2.6-Pro | 69.2 | #37 |
| 12 | MiMo-V2.5 | 66.7 | #192 |
| 13 | MiMo-V2.6-Flash | 65.4 | #62 |
| 14 | Qwen 3.6 Plus | 61.5 | #115 |
| 15 | Qwen 3.8 Max | 61.5 | #31 |
No result here: #7 Gemini 4 Argon, #9 Claude Fable 5.1, #10 Claude Sonnet 5.5.
Interactive version: theaggregate.ai/benchmark?slug=omnistepbench-multimodal-multi-step · How It Works · Data refreshed daily, snapshot 2026-10-16.