Mantis-8B-siglip-llama3: benchmark results
Provider: Salesforce. Released 2024-05-29. Access: Open.
Aggregate Rating 1318 ± 36, rank #8366 of 10916 rated models, from 449 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| OpenVLM MMT-Bench - Spot The Diff | 95 | Score (%) | 97.3 |
| OpenVLM MMT-Bench - Web Shopping | 55 | Score (%) | 94.4 |
| OpenVLM MMT-Bench - Color Constancy | 71.4 | Score (%) | 94.2 |
| OpenVLM MMT-Bench - MIA | 75 | Score (%) | 93.2 |
| OpenVLM MMT-Bench - Scene Emotion Recognition | 65 | Score (%) | 93.2 |
| OpenVLM MMT-Bench - Tech Engineering | 50 | Score (%) | 93 |
| OpenVLM MMT-Bench - Sign Language Recognition | 45 | Score (%) | 91.7 |
| OpenVLM MMT-Bench - Traffic Sign Understanding | 90 | Score (%) | 86.2 |
| OpenVLM Q-Bench - Yes-or-No (In-context Distortion) | 77.8 | Accuracy (%) | 85.8 |
| OpenVLM MMT-Bench - Action Quality Assessment | 30 | Score (%) | 84.7 |
| OpenVLM MMT-Bench - Texture Material Recognition | 80 | Score (%) | 84.5 |
| OpenVLM MMT-Bench - Waste Recognition | 100 | Score (%) | 84.5 |
Interactive version: theaggregate.ai/model?slug=mantis-8b-siglip-llama3 · How It Works · Data refreshed daily, snapshot 2026-11-08.