Zhipu · v4 · depuis 2024-01-16 · 5× · vu le 30 juin 2026
2
Momentum
GLM-4 is a language model series released by Zhipu AI (now Z.ai) in June 2024, consisting of the Open Source GLM-4-9B and the commercial flagship GLM-4-Plus available via the BigModel API. The models support a context window of up to 128K Token (or up to 1M Token in the GLM-4-9B-Chat-1M variant) and master 26 languages. According to Zhipu, GLM-4 achieves performance comparable to or superior to GPT-4 on general Benchmarks such as MMLU, GSM8K, MATH, BBH, GPQA, and HumanEval, and outperforms GPT-4 on the Chinese AlignBench. The models are available partly under an open license (GLM-4-9B) and partly only via the commercial API (GLM-4-Plus).
Historique du momentum
19.05.17.08.
Fonctionnalités
| Key Benchmark (%) | Closely rivals or outperforms GPT-4 on MMLU, GSM8K, MATH, BBH, GPQA and HumanEval (per Zhipu technical report, no exact % given) |
| Context Window (Tokens) | 128,000 tokens standard; up to 1,000,000 tokens in GLM-4-9B-Chat-1M |
| License | GLM-4-9B: open-weight (open-source); GLM-4-Plus: proprietary, API-only |
| Multimodality | Additional vision model GLM-4V-9B (image+text) and voice model GLM-4-Voice (speech) as separate models in the series |
| Platform | API via Zhipu BigModel Open Platform (open.bigmodel.cn); open-weight release for GLM-4-9B via GitHub/Hugging Face |
| Release Date | June 5, 2024 (GLM-4-9B); GLM-4-Plus: August 2024 |