AI
Cerebras has detailed how it spreads 3-trillion-parameter models like Kimi K3 across 59 WSE-3T wafers, keeping weights on-chip and streaming activations between them.
Loading…
Search
No phones and 1 story for “wafer-scale”
AI
Cerebras has detailed how it spreads 3-trillion-parameter models like Kimi K3 across 59 WSE-3T wafers, keeping weights on-chip and streaming activations between them.