InternVL3.5-241B-A28B: benchmark results
Provider: Shanghai AI Lab. Released 2025-08-26. Access: Open.
Aggregate Rating 1631 ± 11, rank #740 of 10916 rated models, from 197 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| LithoBench - Practical Application (Open-Ended) | 3.96 | OPEN_J score (0-10) = 0.4 x ROUGE-L x 10 + 0.2 x BLEU-4 x 10 | 100 |
| MIBench (Multimodal Interaction) - Text-Centric | 74.78 | Accuracy (%) on the text-centric samples (the answer must co | 100 |
| UNICBench (Image) | 21.9 | Exact-count hit rate (%): share of parsed numeric answers eq | 100 |
| OVO-S-Bench - Egocentric Perception | 55.55 | Level 1 (instantaneous egocentric perception) multiple-choic | 96.7 |
| CourtSI-Bench - Player-Zone | 56.1 | Exact-match accuracy (%) on the 82 player-to-court-zone rela | 95.8 |
| MIBench (Multimodal Interaction) | 65.82 | Overall accuracy (%) on the full MIBench (multiple-choice im | 95.2 |
| MIBench (Multimodal Interaction) - Synergy | 58.01 | Accuracy (%) on the synergy samples (both modalities are nee | 95.2 |
| OVO-S-Bench - Context Tracking | 55.72 | Level 2 (spatiotemporal context tracking of evidence no long | 93.3 |
| EPIC-Bench - Affordance Region | 16.85 | Manipulation score (0-100) on affordance region localization | 93.2 |
| EPIC-Bench - Spatial-Related Attributes | 50.59 | Target localization score (0-100) on the spatial-attribute t | 92 |
| LiveClin | 34.4 | Case accuracy (%): share of the 1,407 first-half-2025 clinic | 92 |
| CourtSI-Bench - Camera-Player | 80.2 | Exact-match accuracy (%) on the 248 camera-to-player relatio | 91.3 |
Interactive version: theaggregate.ai/model?slug=internvl3-5-241b-a28b · How It Works · Data refreshed daily, snapshot 2026-11-08.