Cosmos-T2-Accelerate-Preview

9.96M params 4 layers RoPE + RMSNorm + SwiGLU + GQA Engram on PREVIEW CPU

Trained on wop/XXXXXL-chain-of-thought  ·  Model repo

⚠️ Preview / research model — tiny (~10M params), will hallucinate freely and locks into the <think>…</think> Answer template. Keep temperature low for stable outputs.
Send a message to start…
0 2
1 200
64 1028
16 1028