BREAKING
Z.ai releases GLM-5.2 open-weight
0tok
Context window
0x
vs GLM-5.1
0%
DeepSWE score
Two reasoning modes
maxPerformance
Highest reasoning effort
For hard tasks
highEfficiency
Balanced effort
Faster and cheaper
DeepSWE gap to proprietary
GLM-5.246.2
GPT-5.570
MIT weights, unchanged pricing
1Hugging Face zai-org
2Z.ai API & OpenRouter
31.5TB, quantize local
Open-weight race heats up
AI NEWS BLITZ
Chinese lab Z.ai has launched GLM-5.2, a new open-weight model for coding and agentic tasks.