56 6 days ago

Tencent Hunyuan-A13B 80.4B Sparse MoE (13B active, 256K ctx, Q4_K_M, THINK_LEN=0) — 100% SecOps/Red-Team compliance (8/9 overall), 16.7% Hallucination (5/6), 151.0 tok/s decode & 3,992 tok/s prefill in 46.8 GiB VRAM.

80b