speculative-decoding 4

• Turbo3 + MTP: Merging Two llama.cpp Forks May 15, 2026
• Qwen 3.6 Dense vs MOE on Local Stack: what MTP actually delivers May 13, 2026
• Qwen 3.6 27B with Native MTP on llama.cpp May 13, 2026
• 1.5× Faster Agentic Coding with MTP on Qwen 3.6 27B May 8, 2026