Decode · tok/s
57
Prefill
—
tok/100W
22.8
Decode 57 tok/s. qwen3.8-flash-next-unsloth-ud-iq4_xs. context window 131K. VRAM Tesla V100 32GB. GPU電力制限 150v.
First-person V100 32GB. Rank is the quoted 57 decode. No prefill in the post. Separate from the 2× V100 pair.
