BREAKING
Zhipu Weighs Custom AI Chip for GLM
0
x
daily token usage jump
0
%
related HK shares low
0
%
related HK shares high
Why Build In-House Silicon
1
Cut inference cost
↓
2
Secure compute
↓
3
Reduce NVIDIA reliance
↓
4
Full-stack integration
GLM-5.2 vs Rivals on Benchmarks
GLM Term-Bench
81
Opus 4.8
85
GLM SWE-Pro
62.1
State of the Plan
Confirmed
●
Early talks with chip firms
●
GLM inference ASIC focus
●
Optimized for domestic hardware
Unknown
●
No partners disclosed
●
No specs or timeline
●
Could take over two years
China's Full-Stack AI Ambition
AI NEWS BLITZ
Chinese AI lab Zhipu is in early talks to design its own chip for GLM models.