InternVL3.5-38B-Instruct: benchmark results
Provider: Shanghai AI Lab. Released 2025-08-26. Access: Open.
Aggregate Rating 1581 ± 25, rank #1246 of 10916 rated models, from 26 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| MCSBench - L1 | 71.26 | Accuracy (%) | 75.7 |
| MCSBench - L2 | 60.05 | Accuracy (%) | 75 |
| MCSBench - Overall | 54.46 | Accuracy (%) | 75 |
| Omni-I2C (SMILES) - Tanimoto Similarity | 58.8 | Tanimoto similarity (0-100) between Morgan fingerprints of t | 66.7 |
| Video-IFBench - Nested | 8.4 | Task-gated instruction satisfaction rate (%; TISR on nested | 59.5 |
| MCSBench - L3 | 24.43 | Accuracy (%) | 58.6 |
| Video-IFBench - Single | 24.5 | Task-gated instruction satisfaction rate (%; TISR on single- | 48.6 |
| K-MetBench | 57.3 | Accuracy (imported) | 48.3 |
| Video-IFBench | 20.8 | Task-gated instruction satisfaction rate (%; TISR over all f | 43.2 |
| Video-IFBench - Multi | 27.1 | Task-gated instruction satisfaction rate (%; TISR on multi-t | 37.8 |
| Omni-I2C (LaTeX) - Visual Fidelity | 38.6 | Visual fidelity S_img (0-100): GPT-4.1 ratings of compositio | 33.3 |
| Omni-I2C (Python) - Visual Fidelity | 59.2 | Visual fidelity S_img (0-100): GPT-4.1 ratings of compositio | 33.3 |
Interactive version: theaggregate.ai/model?slug=internvl3-5-38b-instruct · How It Works · Data refreshed daily, snapshot 2026-11-08.