GLM-4.5V
106B-total (12B active) open vision-reasoning model with a thinking-mode toggle, reporting state of the art among same-size models on 42 benchmarks.
Handles images, video, documents and GUI operation. The thinking toggle lets users trade latency for deeper reasoning.
- Date
- Monday, 11 August 2025
- Lab
- Zhipu AI / Z.ai
- Kind
- open-weights
- Access
- open weights
Figures
| Measure | Value | Measured by |
|---|---|---|
| Public VLM benchmarks with SOTA at its scale | 42 company-claimed | company |
Benchmarks are company-claimed.
Sources
This record was checked against its sources on 6 October 2026. How we check