Tag
Benchmark results for running Qwen 3.6 27B on AMD MI50 GPUs using a custom vllm fork, achieving 52.8 tokens/s TG and 1569 tokens/s PP without quantization or MTP, demonstrating usability for agentic tasks on 2018 hardware.
A developer reports that the new 27B Qwen 3.6 model runs excellently on a 24GB VRAM laptop, passing all PySpark/Python data-transformation benchmarks and eliminating the need for cloud subscriptions.
Qwen 3.6 35B achieves near-perfect 283/285 line recall on a 108 k-token JS file, outperforming Gemma 4 27B (6/16 passes) and fixing long-context weaknesses of earlier Qwen versions.