Idefics3: benchmark results
Provider: HuggingFace. Access: Open.
Aggregate Rating 1309 ± 35, rank #8565 of 10916 rated models, from 545 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| MEGA-Bench Task - Constrained Generation Multi Contain Position Only | 26.7 | Task Score (%) | 98.1 |
| MEGA-Bench Task - Video Grounding Temporal | 40 | Task Score (%) | 94.2 |
| MEGA-Bench Task - Red Teaming Celebrity | 88 | Task Score (%) | 88.4 |
| MEGA-Bench Task - Symbolic Graphics Programs Scalable Vector Graphics | 16.7 | Task Score (%) | 88.4 |
| MEGA-Bench Task - IconQA | 47.4 | Task Score (%) | 86.5 |
| MEGA-Bench Task - ASCII Art 30 | 28.6 | Task Score (%) | 86 |
| MEGA-Bench Task - Geometry Analytic | 21.4 | Task Score (%) | 82.6 |
| MEGA-Bench Task - Music Sheet Sentiment | 42.9 | Task Score (%) | 80.2 |
| MEGA-Bench Task - Google Street View Circle Reasoning | 26.7 | Task Score (%) | 79.1 |
| MEGA-Bench Task - Music Sheet Name | 33.3 | Task Score (%) | 79.1 |
| MEGA-Bench Task - Counting | 64.3 | Task Score (%) | 78.8 |
| MEGA-Bench Task - Planning Visual Storage | 6.7 | Task Score (%) | 77.9 |
Interactive version: theaggregate.ai/model?slug=idefics3 · How It Works · Data refreshed daily, snapshot 2026-11-08.