ChatGLM2 Models by Tsinghua Knowledge Engineering Group (THUDM)
Tsinghua Knowledge Engineering Group (THUDM) releases · 1 in the last 12 monthsChangelog →
This model family is considered obsolete. Consider newer alternatives in Related Model Families below.
1 model2023Up to 32K ctx
Last refreshed 2026-05-01. Next refresh: weekly.
About
The second generation of THUDM's open-source bilingual (Chinese-English) ChatGLM model family. ChatGLM2-6B improved significantly over v1 with a 32K context window (extended via FlashAttention), 1.4T training tokens, and approximately 23% better MMLU performance. Released June 2023. Succeeded by ChatGLM3 in October 2023.
Decision facts
- Best fit
- General model comparison
- Capability starting point
- ChatGLM2-6B with 32K context
- Lowest tracked input
- Not tracked
- Closest related family
- ChatGLM-4
Current Variants
Use-when guidance is based on each model's tracked capabilities, context window, release date, and replacement status.
1 in view
ChatGLM2-6BCurrent
Use when the workload needs 32K context and 6.2B parameters.
2023-0632K context6.2B parameters
| Model | Use when | Released | Signals | Status |
|---|---|---|---|---|
| ChatGLM2-6B | Use when the workload needs 32K context and 6.2B parameters. | 2023-06 | 32K context6.2B parameters | Current |
Release Timeline
1 release group2023-06
1 current
ChatGLM2-6B
Current32K context6.2B parameters
Specifications(1 models)
| Model | Released | Context | Parameters |
|---|---|---|---|
| ChatGLM2-6B | 2023-06 | 32K | 6.2B |
