Qwen3.6-35B-A3B-APEX Runs 128K Context on RTX 3060 12GB

Qwen3.6-35B-A3B-APEX / 128K ctx on RTX 3060 12GB — 37 t/s gen with 72k ctx filled, PPL 3.25, offloading 17GB model

old-mike ran mudler/Qwen3.6-35B-A3B-APEX-MTP-I-Compact.gguf through spiritbuun’s llama.cpp fork on one RTX 3060 12GB, offloading a 17.3GB model and reaching 37.17 t/s generation at 72K filled context, 28.08 t/s at 129K, and PPL 3.2529 on an enwik8 64K-context perplexity test.

Why it matters: HKR-H/K/R all pass via a concrete consumer-GPU inference result with speed and PPL. Source is a single Reddit post and the impact stays within local inference, so it lands in featured, not P1.