LLMOpen weights
GLM-4.6
Zhipu AI
Updated May 21, 2026 · Benchmarks via Artificial Analysis, specs & pricing via OpenRouter · Methodology · Report an error
Superseded by GLM-5
Specifications
- Type
- LLM
- Access
- Open weights
- Released
- September 30, 2025
- License
- MIT
- Parameters
- 357B (MoE)
- Context window
- 202,752 tokens
- Knowledge cutoff
- March 31, 2025
- Output speed
- 85 tok/s
- Latency (TTFT)
- 0.7s
- Input
- Text
- Output
- Text
- API pricing
- $0.43 in · $1.74 out / 1M tokens
Capabilities
Function callingStructured output
Benchmarks
Our take
Expanded context to 200K and improved coding efficiency (~15% fewer tokens than 4.5), reaching near-parity with Claude Sonnet 4 on real-world coding.