60-70 tokens/sec Windows 11 LLM Studio 250 or whatever the models max is below that as context Qwen3.6 35B A3B 35B-A3B, gguf