ChatGLM2-6B
Second-generation 6B open model with context extended from 2K to 32K tokens and 42% faster inference.
Upgraded base and chat model. Zhipu reports +23% on MMLU and +33% on C-Eval over v1, context raised from 2K to 32K, and 42% faster inference. It also underpinned CodeGeeX2 (2023-07-25).
- Date
- Sunday, 25 June 2023
- Lab
- Zhipu AI / Z.ai
- Kind
- open-weights
- Access
- open weights (restricted license)
Figures
| Measure | Value | Measured by |
|---|---|---|
| MMLU improvement vs ChatGLM-6B | +23% company-reported | company |
| C-Eval improvement vs ChatGLM-6B | +33% company-reported | company |
| Context length | 2K to 32K | company |
Date and numbers from the README news log of ChatGLM-6B; benchmark gains are company-reported.
Sources
This record was checked against its sources on 6 October 2026. How we check