Community benchmarks compare Qwen 3.6 speeds on multi-GPU setups

Community benchmarks compare Qwen 3.6 on multi-GPU setups. A 4×RTX 5060 Ti (16GB) runs 27B FP8 at 55 tok/s with 262K context ($1800 total). On dual Radeon R9700, vLLM beats ROCm and Vulkan for speed.