@UnslothAI Doing some comparison between unsloth/Qwen3.6-27B-UD-IQ3_XXS.gguf vs Claude Sonnet 4.6
With the same prompt creating a single file html simulation, in my opinion the, Qwen3.6 got better and beautiful result
Here's how i manage running Qwen 3.6 27B smoothly on my RTX 5060 Ti 16GB! Stable inference at ~45 tokens/sec average with full 128K context
Using:unsloth/Qwen3.6-27B-MTP-GGUF
llama.cpp mtp-clean branch
Big thanks to @unslothAI and @am17an for the MTP