GLM-4.7-Flash 31B-A3B
#17 en Modèles de langage open sourceZhipu Ai · v4.7 · flash 31b a3b · depuis 19. Januar 2026 · 4× · vu le 14 oct. 2026
45
Momentum
GLM-4.7-Flash is an open-weight language model from Z.ai (formerly Zhipu AI) with a 30B-A3B Mixture-of-Experts architecture (about 31B total parameters, about 3B active per token). It is positioned as the lightweight variant of the GLM-4.7 series and is offered free via the Z.ai API and as open weights for self-hosting. The benchmark values come from the vendor's Hugging Face model card.
Historique du momentum
20.07.18.10.
Fonctionnalités
| Key Benchmark (%) | SWE-bench Verified: 59.2% |
| Context Window (Tokens) | 200K (202,752 tokens per Unsloth docs); up to 128K output tokens |
| License | MIT |
| Multimodality | Text only (text input and output) |
| Platform | Z.ai API (model ID glm-4.7-flash); open weights on Hugging Face, runs locally (vLLM, SGLang) |
| Price per 1M Tokens | Z.ai API: free (input, cached input, output = $0) |
| Release Date | January 19, 2026 |