Hybrid-reasoning GLM release that made the 4.5 line broadly useful
Capabilities
ReasoningYes
Tool callingYes
Structured outputNo
TemperatureSupported
AttachmentNo
Open weightsYes
Limits
Context131,072 tokens
Max inputN/A
Max output98,304 tokens
Details
Familyglm
Knowledge cutoff2025-04
Release date2025-07-28
Last updated2025-07-28
Benchmarks
Artificial Analysis Coding Index
26.3
index
SciCode
34.8
percent correct
Terminal-Bench Hard
22
success rate
Comments (0)
Sign in to leave a comment
No comments yet.