whatmodel.sh / ChatGLM2-6B
ChatGLM2-6B
↗ share
+ contribute
ChatGLM2-6B is an open-weights language model from Zhipu AI (THUDM), released June 2023. Its published size is 6B parameters. Licence: chatglm2-6b-license.
Catalogue identity notes
Comparison-only. Arena entry 'chatglm2-6b' (text). LMArena's leaderboard table (arena-leaderboard space @6f2eeeb) labels 'chatglm2-6b' as 'ChatGLM2-6B' (Tsinghua, Apache-2.0), linking https://huggingface.co/THUDM/chatglm2-6b. Maker repo zai-org/chatglm2-6b @d2e2d91: licence tag none tagged. Parameters: nominal size from the model name. Released 2023-06: month the maker's Hugging Face repo was created (createdAt 2023-06-24).
Model information sources
$ how it compares
The LLMs ranked either side of ChatGLM2-6B, out of 828 with a score, then more from Zhipu AI (THUDM).
| # | model | score | where | ||
|---|---|---|---|---|---|
| 777 |  Pythia 12B | 25±1 | local | ||
| 778 |  RedPajama-INCITE Base 3B v1 | 25±3 | local | ||
| 779 | ChatGLM2-6B | 25±2 | local | ||
| 780 | GPT-J 6B | 25±1 | local | ||
| 781 |  OpenLLaMA 3B v2 | 25±1 | local | ||
| more from Zhipu AI (THUDM) | |||||
| 349 | GLM-4 9B Chat 1M | 53±5 | local | ||
| 485 | CogVLM2 Llama3 Chinese Chat 19B | 47±7 | comparison only | ||
| 569 | CogVLM2 Llama3 Chat 19B | 43±3 | local | ||
$ what it's good at
chat & writing▁▃▁▃▅▄▅█▆▆▇▃ #387 of 393 · 96% of its score · 5 tests
Beats Phi-2 (2.7B) here, though they rank higher overall. Trails Gemma 3 270M IT, SmolLM2 360M Instruct here, though they rank lower overall.
| test | result | vs the others in its run | counts |
|---|---|---|---|
| Polyrating rating (bias-aware re-fit of Chatbot Arena votes, mid-2024) | 930.0 ✎ | ▂▁▃▁▃▇█▆▇▄▂▂ beats 7/95 | 23% |
| LMArena (text) | 1024.7 ✎ | ▁▃▁▂▃▃▃▆▆▇█▃ beats 6/379 | 23% |
| MT-Bench (GPT-4 single-answer grading, LMSYS runs) | 4.9600 MT-Bench score (1-10) ✎ | ▆▂▂▆▆▄██▂▄▄▆ beats 8/28 | 20% |
| AlpacaEval 2 length-controlled | 4.4% ✎ | ▁▃▂▃▆▆█▆▄▄▁▁ beats 5/122 | 15% |
| AlpacaEval 1.0 win rate (vs text-davinci-003, GPT-4 judge) | 47.1% ✎ | ▂▂·▆·▇▃▆█▆▄▂ beats 7/45 | 14% |
other tests1 not in a task · 2.8% of its score
| test | result | vs the others in its run | counts |
|---|---|---|---|
| Open PT-LLM Leaderboard average (9 Portuguese tasks) | 47.2% ✎ | ▃▁▁▁▁▁▁▃▂█▅▁ beats 62/276 | 2.8% |
$ where to run it
Other ways to run it
- Run it as published · needs ~12.49 GB, NVIDIA only · from github.com/vllm-project
vllm serve zai-org/chatglm2-6b --trust-remote-code
Weights and documentation: huggingface.co/zai-org ↗
$ how the score was worked out · AlpacaEval 1.0 and 2.0 leader… 29%, Polyrating: A Cost-Effective… 23%, LMArena leaderboard dataset 23% and 2 more
Each result implies a score on its own, after adjusting for how hard its test is and who ran it. The score is their weighted average, with a small pull toward the middle for thinly tested models. Each test's weight is its counts column above; here it's added up by who ran the tests. How it works
| source | results | share of the score |
|---|---|---|
 AlpacaEval 1.0 and 2.0 leaderboards (tatsu-lab/alpaca_eval repository) | 2 | 29% |
 Polyrating: A Cost-Effective and Bias-Aware Rating System for LLM Evaluation (arXiv:2409.00696v3) | 1 | 23% |
 LMArena leaderboard dataset | 1 | 23% |
 LMSYS Chatbot Arena leaderboard Space: MT-Bench column of the published leaderboard table | 1 | 20% |
| Open Portuguese LLM Leaderboard (archived) | 1 | 3% |
| prior · pull toward the middle | 1% | |
| weighted average | = 24.8 |
$ improve this page
Use a short GitHub form or edit the data through a pull request. Every number needs a source; here's which ones we can use.
data updated · sources and licences · json