1 10 hours ago

100% Uncensored (9/9 Suite A) InclusionAI Ring-Flash-2.0 102.9B Reasoning MoE (6.1B active, 32K ctx, i1-Q3_K_M) — 286.0 tok/s decode & 6,090 tok/s prefill in 46.8 GiB VRAM on H100 80GB (THINK_LEN=0 + CoT guide).

103b