

GLM-4.7
#291 in Frontier LLMsZhipu · v4.7 · since 22. Dezember 2025 · 8× · last seen Jun 30, 2026
GLM-4.7 is an open-source frontier language model released by Zhipu AI (Z.ai) on December 22, 2025, using a Mixture-of-Experts architecture (358 billion total parameters, roughly 32 billion active). It is optimized for agentic coding, tool use, and multi-step reasoning, offering a context window of about 200,000 tokens and an output capacity of up to 128,000–131,000 tokens. It is freely available under the MIT license (Hugging Face, ModelScope) and achieves 73.8% on SWE-bench Verified and 42.8% on the Humanity's Last Exam benchmark. Via the Z.ai API it is priced at roughly $0.40–$0.60 per million input tokens and $1.75–$2.20 per million output tokens.
Features
| Key Benchmark (%) | SWE-bench Verified: 73.8% (+5.8% vs GLM-4.6) |
| Context Window (Tokens) | approx. 200,000 tokens (202,752), max output 131,072 tokens |
| License | MIT License (open source, commercial use permitted) |
| Multimodality | Primarily text-based (multimodal input per third-party listing); separate vision variant GLM-V/GLM-4.6V available |
| Platform | Z.ai API/Open Platform, chat.z.ai, Hugging Face, ModelScope; compatible with vLLM and SGLang |
| Price per 1M Tokens | API (official): $0.60 input / $2.20 output per 1M tokens |
| Release Date | December 22, 2025 |