LLMOpen weights
GLM-4.5
Zhipu AI
Updated May 21, 2026 · Benchmarks via Artificial Analysis, specs & pricing via OpenRouter · Methodology · Report an error
Superseded by GLM-4.6
Specifications
- Type
- LLM
- Access
- Open weights
- Released
- July 28, 2025
- License
- MIT
- Parameters
- 355B (32B active)
- Context window
- 131,072 tokens
- Knowledge cutoff
- December 31, 2024
- Output speed
- 85 tok/s
- Latency (TTFT)
- 0.7s
- Input
- Text
- Output
- Text
- API pricing
- $0.60 in · $2.20 out / 1M tokens
Capabilities
Function callingStructured output
Benchmarks
Reasoning
| GPQA Diamond | 79.1 |
Coding
| LiveCodeBench | 72.9 |
| SWE-bench Verified | 64.2 |
| Terminal-Bench | 37.5 |
Agents
| TAU-bench Retail | 79.7 |
| TAU-bench Airline | 60.4 |
| BrowseComp | 26.4 |
General
| MMLU-Pro | 84.6 |
| Humanity’s Last Exam | 14.4 |
Our take
Zhipu’s first 355B agentic/reasoning/coding flagship under MIT, with a hybrid thinking/direct mode and a smaller Air variant.