I developed a game using Qwen3.6-27B-TQ3_4S on 5060ti 16GB RAM fully loaded 32k context

Use two prompts: 1st instruction. It didn't start. Then paste the error. It fixed it and it works! Decode speed about 24-26 tok/s. Processing Speed 650 to 700 tok/s. It take longer than A35B but it create more workable solution in 2 shots. https://huggingface.co/YTan2000/Qwen3.5-27B-TQ3_1S Need private llama fork https://github.com/turbo-tan/llama.cpp-tq3 If you want to use it in LMStudio, follow instruction https://github.com/turbo-tan/llama.cpp-tq3/blob/main/docs/backend/LMStudio.md