SenseNova-U1-8B-MoT: benchmark results
Provider: SenseTime. Released 2026-04-27. Access: Open.
Aggregate Rating 1532 ± 47, rank #1827 of 10546 rated models, from 18 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| WeGenBench - Text Rendering | 74 | Character-level F1 (0-1 scaled to %) between the text render | 82.4 |
| WeGenBench - Alignment | 93 | Checklist-based QA score (0-1 scaled to %): a Qwen3.5-397B-A | 58.8 |
| Game2Real-Bench - Real World - Spatial Reasoning | 33 | Task score (%) | 58.3 |
| Game2Real-Bench - Toy Game - Spatial Reasoning | 16.7 | Accuracy (%) | 52.1 |
| WeGenBench - Alignment (COT Review) | 7.81 | COT-based deduction score (1-10): a fine-tuned expert vision | 47.1 |
| Game2Real-Bench - Toy Game - Sequential Planning | 18 | Accuracy (%) | 39.6 |
| Game2Real-Bench - Real World - Sequential Planning | 36 | Task score (%) | 37.5 |
| Game2Real-Bench - Toy Game - Fine-grained Perception | 13.5 | Accuracy (%) | 37.5 |
| WeGenBench - Text Rendering Sentence Accuracy | 52 | Sentence accuracy (0-1 scaled to %), the exact-match rate of | 35.3 |
| GameHorizon - Offline T2 Multi-Horizon Decomposition | 59.2 | Accuracy (%) | 33.3 |
| GameHorizon - Offline T3 Cross-Horizon Consistency | 67.5 | Accuracy (%) | 28.6 |
| Game2Real-Bench - Overall | 18.5 | Mean per-sample score (%) | 25 |
Interactive version: theaggregate.ai/model?slug=sensenova-u1-8b-mot · How It Works · Data refreshed daily, snapshot 2026-11-18.