跳转到主要内容

GLM-5.1 (Reasoning)

Z AIGLMOpen WeightMIT · Commercial OK

描述

GLM-5.1 is Z.AI's next-generation flagship foundation model designed for long-horizon agentic engineering tasks. Built on a 754B MoE architecture (40B active parameters), it can work continuously and autonomously on a single task for up to 8 hours, completing the full loop from planning and execution to iterative optimization and delivery. GLM-5.1 achieves state-of-the-art on SWE-Bench Pro (58.4) and demonstrates strong performance across coding, reasoning, and agentic benchmarks. It supports 200K context length, 128K max output tokens, thinking mode, function calling, structured output, context caching, and MCP integration. Overall performance is aligned with Claude Opus 4.6 with particular strengths in sustained execution and complex engineering optimization.

发布日期
2026-04-07
参数规模
754.0B
上下文长度
203K
支持模态
text

能力雷达图

46
general
43
coding
87
reasoning
60
science估算
60
agents
0
multimodal

Science 在缺少专门科学评测时使用推理能力代理估算。

排行榜排名

领域#排名分数来源
智能体与工具21
67.0
LS
代码能力榜40
75.0
AA
通用能力榜9
90.0
AA
科学能力33
76.0
AA

基准测试分数 (LLM Stats)

Agents

Vending-Bench 2563441.0%自报
BrowseComp79.3%自报
MCP Atlas71.8%自报
TAU3-Bench70.6%自报
Terminal-Bench 2.069.0%自报
CyberGym68.7%自报
SWE-Bench Pro58.4%自报
NL2Repo42.7%自报
Toolathlon40.7%自报

Biology

GPQA86.2%自报

Math

AIME 202695.3%自报
HMMT 202594.0%自报
IMO-AnswerBench83.8%自报
HMMT Feb 2682.6%自报
Humanity's Last Exam52.3%自报

AA 评测指数

Intelligence Index
51.4
Coding Index
43.4
Tau2
1.0
Gpqa
0.9
Ifbench
0.8
Lcr
0.6
Scicode
0.4
Terminalbench Hard
0.4
Hle
0.3

LLM Stats 分类评分

Agents
100
Reasoning
100
Biology
90
Chemistry
90
General
90
Physics
90
Math
80
Search
80
Code
70
Safety
70
Tool Calling
60
Vision
50
Coding
40

定价

输入价格$1.4 / 1M tokens
输出价格$4.4 / 1M tokens
混合价格(3:1)$2.15 / 1M tokens

速度

Tokens/秒53.8 tokens/s
首Token延迟1.04s
首回答延迟71.55s

可用提供商

(LS 内部计价单位)
提供商输入价格输出价格
ZAI1.4M4.4M

外部链接