EgoBench - Dynamic Easy Mode - Micro Accuracy: leaderboard
Egocentric multimodal tool use (EgoBench, Liu et al., 2026): an agent watches a short first-person video and serves a simulated customer in retail, restaurant, food-ordering and kitchen scenarios by calling scenario tools over a database, with thinking disabled, across 1,045 tasks. Here the simulated user reveals the request one step per turn and cooperates. MicroAcc is the percent of all ground-truth tool calls, pooled over tasks, that the agent issued, with no penalty for extra calls; higher is better.
Metric: MicroAcc (%; share of ground-truth tool calls issued, pooled over tasks, cooperative simulated user revealing one step per turn). Source: arxiv.org. Saturation forecast: Around February 2027. 8 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Gemini 3.1 Pro (Preview) | 44.42 | #91 |
| 2 | Qwen 3.5 397B A17B (Non-reasoning) | 41.66 | #242 (Qwen 3.5 397B A17B) |
| 3 | Qwen 3.6 Plus (Non-reasoning) | 36.58 | #217 (Qwen 3.6 Plus) |
| 4 | Kimi K2.5 (Non-reasoning) | 35.42 | #252 (Kimi K2.5) |
| 5 | GLM-5V Turbo | 29.7 | #342 |
| 6 | Qwen 3 VL 235B A22B Instruct | 15.02 | #521 |
No result here: #1 Claude Fable 5.2, #2 Gemini 4 Carbon, #5 Claude Opus 5.5.
Interactive version: theaggregate.ai/benchmark?slug=egobench-dynamic-easy-mode-micro-accuracy · How It Works · Data refreshed daily, snapshot 2026-11-12.