GLM-4.5V (Non-reasoning): benchmark results
Provider: Zhipu. Released 2025-08-11. Access: Open.
Aggregate Rating 1546 ± 15, rank #912 of 2205 rated models, from 55 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| CFMME - Seal Detection | 37.45 | Seal detection mAP (%): boxes and categories (company or per | 100 |
| CFMME - Information Extraction | 89.03 | Field-level F1 (%) of key information extracted from financi | 87.5 |
| CFMME - Recognition, Detection and Extraction | 71.77 | Average (%) of seal recognition 1-NED, table recognition TED | 87.5 |
| CFMME - Table Recognition | 71.89 | Table recognition TEDS (%, tree-edit-distance similarity of | 87.5 |
| CFMME - Application QA | 50.13 | Accuracy (%) on question answering over real financial docum | 81.2 |
| AA Omniscience - Software Engineering (SWE) - JavaScript | 37.27 | Accuracy (%) | 75.8 |
| CFMME - Knowledge | 64.9 | Accuracy (%) on the knowledge assessment (financial subject | 75 |
| M-JudgeBench - Short CoT Cross-Model Pairs | 69.02 | pairwise accuracy (%) on the 297 pairs where a correct and a | 68.4 |
| M-JudgeBench - Visual Perception Error Detection | 93.15 | pairwise accuracy (%) on the 248 pairs where the rejected ch | 68.4 |
| AA Omniscience - Software Engineering (SWE) - PHP | 26 | Accuracy (%) | 65 |
| GeoAux-Bench (Core) - Olympiad Junior | 82.61 | Accuracy (%) on the 207 junior olympiad questions of GeoAux- | 65 |
| GeoAux-Bench (Core) - Olympiad Senior | 72.15 | Accuracy (%) on the 237 senior olympiad questions of GeoAux- | 65 |
Interactive version: theaggregate.ai/model?slug=glm-4-5v-non-reasoning · How It Works · Data refreshed daily, snapshot 2026-11-13.