wuhp/Qwen3.6-35B-A3B-MTP-GGUFtest

Qwen3.6-35B-A3B served via llama.cpp (b9222) with Multi-Token Prediction speculative decoding.

• —Model: unsloth/Qwen3.6-35B-A3B-MTP-GGUF @ UD-Q2KXL (~12.6 GB)
• —MTP: --spec-type draft-mtp --spec-draft-n-max 2
• —KV cache: q80 keys / iq4nl values
• —Context: 4096 tokens
• —Hardware: CPU Basic (2 vCPU / 16 GB RAM)

OpenAI-compatible API on port 7860.