BREAKING
GLM-5.2 Tested vs Claude Opus
"A real frontier coding model"
0B
active params
0B
total params
0M
context tokens
Terminal-Bench 2.1 scores
GLM-5.162
GLM-5.281
Opus 4.885
GLM-5.2 vs Opus tradeoffs
GLM-5.2open
Free local execution
Strong tool-calling
1M context agentic work
Claude Opuspremium
Better intent understanding
One-shot accuracy
Roughly 10x the price
Open weights, MIT license
AI NEWS BLITZ
An open-weight coding model from Z.ai goes head to head with Claude Opus, running locally for free.