InternVL3.5-14B-Instruct: benchmark results
Provider: Shanghai AI Lab. Released 2025-08-26. Access: Open.
Aggregate Rating 1548 ± 24, rank #1625 of 10870 rated models, from 52 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| SiT-Bench - Global Perception & Mapping | 23.64 | Accuracy (%; item-weighted over the category's subtasks) | 76.7 |
| SiT-Bench - Logic & Anomaly Detection | 37.56 | Accuracy (%; item-weighted over the category's subtasks) | 76.7 |
| SiT-Bench - Navigation & Planning | 49.33 | Accuracy (%; item-weighted over the category's subtasks) | 76.7 |
| MCSBench - L3 | 26.72 | Accuracy (%) | 76.3 |
| CulMind - Ancient Scripts | 6.8 | Ancient Scripts subdomain score: macro-average of the task s | 69.2 |
| SiT-Bench | 41.47 | Accuracy (%; item-weighted over the 17 subtasks) | 69 |
| EgoPoint-Bench - Real-World Function & State | 65.09 | Accuracy (%) on the Function & State questions among the 1,1 | 63.6 |
| SiT-Bench - Embodied & Fine-grained Perception | 42.81 | Accuracy (%; item-weighted over the category's subtasks) | 63.3 |
| SiT-Bench - Multi-View & Geometric Reasoning | 46.17 | Accuracy (%; item-weighted over the category's subtasks) | 61.7 |
| MCSBench - L2 | 53.27 | Accuracy (%) | 60.5 |
| SpatiaLQA - Content F1 | 56.8 | Step-content F1 (%, 0-100): harmonic mean of the pooled reca | 60 |
| MCSBench - Overall | 48.41 | Accuracy (%) | 59.2 |
Interactive version: theaggregate.ai/model?slug=internvl3-5-14b-instruct · How It Works · Data refreshed daily, snapshot 2026-11-12.