GLM-4.5 is an Agentic, Reasoning, and Coding (ARC) foundation model designed for intelligent agents, featuring 355 billion total parameters with 32 billion active parameters using MoE architecture. Trained on 23T tokens through multi-stage training, it is a hybrid reasoning model that provides two modes: thinking mode for complex reasoning and tool usage, and non-thinking mode for immediate responses. The model unifies agentic, reasoning, and coding capabilities with 128K context length support. It achieves exceptional performance with a score of 63.2 across 12 industry-standard benchmarks, placing 3rd among all proprietary and open-source models. Released under MIT open-source license allowing commercial use and secondary development.
Benchmarks
评测成绩
| 评测基准 | 类别 | 分数 | 来源 |
|---|---|---|---|
| MATH-500 | math reasoning | 98.2 | 来源 |
| AIME24 | math reasoning | 91.0 | 来源 |
| MMLU-Pro | language reasoning math general | 84.6 | 来源 |
| TAU-bench-Retail | reasoning communication | 79.7 | 来源 |
| GPQA | reasoning general | 79.1 | 来源 |
| BFCL-v3 | general reasoning | 77.8 | 来源 |
| LiveCodeBench | reasoning general code | 72.9 | 来源 |
| AA-Index | general | 67.7 | 来源 |
| SWE-bench-Verified | reasoning frontend_development code | 64.2 | 来源 |
| TAU-bench-Airline | reasoning communication | 60.4 | 来源 |
| SciCode | reasoning math physics chemistry code | 41.7 | 来源 |
| Terminal-Bench | reasoning code | 37.5 | 来源 |
| BrowseComp | reasoning search | 26.4 | 来源 |
| HLE | reasoning math | 14.4 | 来源 |
Pricing
API 价格对比
| 服务商 | 输入价 | 输出价 | 上下文 | 吞吐(tok/s) | 延迟(s) | 函数调用 | 代码执行 | 联网搜索 |
|---|---|---|---|---|---|---|---|---|
| 302.AI | $0.29 | $1.14 | 131K | — | — | ✓ | ✗ | ✗ |
| LLM Gateway | $0.60 | $2.20 | 131K | — | — | ✓ | ✗ | ✗ |
| Z.AI | $0.60 | $2.20 | 131K | — | — | ✓ | ✗ | ✗ |
| Zhipu AI | $0.60 | $2.20 | 131K | — | — | ✓ | ✗ | ✗ |
| Cortecs | $0.67 | $2.46 | 131K | — | — | ✓ | ✗ | ✗ |
价格单位：美元/百万 token，数据来自社区整理，仅供参考。