General GLM flagship for coding, analysis, and tool-heavy engineering workflows
Capabilities
ReasoningYes
Tool callingYes
Structured outputNo
TemperatureSupported
AttachmentNo
Open weightsYes
Limits
Context204,800 tokens
Max inputN/A
Max output131,072 tokens
Details
Familyglm
Knowledge cutoffN/A
Release date2026-02-12
Last updated2026-02-12
Benchmarks
SWE-Bench Verified
72.8
resolved
SWE-Atlas Codebase QnA
20.5
score
SWE-Atlas Refactoring
24.24
score
SWE-Atlas Test Writing
28.74
score
Comments (0)
Sign in to leave a comment
No comments yet.