GLM-5V Turbo: benchmark results
Zhipu's GLM-5V Turbo native multimodal vision-coding model for design-to-code and agentic workloads. Provider: Zhipu. Released 2026-04-01. Access: API.
Aggregate Rating 1717 ± 8, rank #166 of 1656 rated models, from 283 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| LDU-Bench - Morphology Recognition | 40.8 | Macro-F1 (%; normal plus 11 defect morphology labels on the | 100 |
| LLM Stats (SimpleVQA) | 78.2 | Score (%) | 100 |
| PAWBench (VLM Prediction) - Calibration | 34.8 | Conditional total-variation distance for vision-language mod | 100 |
| Tau2-Bench Telecom | 98.54 | Success Rate (%) | 99.2 |
| AA TAU-2 Bench | 98.54 | Accuracy (%) | 99 |
| Horangi 4 - BFCL | 70.16 | Accuracy (%) | 95 |
| CATTest - Pass/Not-Pass Precision | 86.79 | Precision (%) | 93.8 |
| MatPhaseBench | 0.4 | BERTScore recall (XLNet-large semantic recall, mean over 200 | 91.7 |
| TOC-Bench - Duration Category | 42.3 | Duration Category accuracy (%) on TOC-Bench (2,323 human-ver | 90 |
| VideoGAIA - Culture | 51.72 | Pass@1 accuracy (%; the 58 culture tasks; unified ReAct agen | 86.8 |
| SuperCLUE-AdsVU - Placement Feasibility | 90.35 | Score | 85.7 |
| TOC-Bench - Reappearance or Disappearance | 67.9 | Reappearance or Disappearance accuracy (%) on TOC-Bench (2,3 | 85 |
Interactive version: theaggregate.ai/model?slug=glm-5v-turbo · How It Works · Data refreshed daily, snapshot 2026-10-20.