Qwen 3.8 27B in full BF16 on one 96GB M3 Ultra Mac Studio.
54.74GB weights • 58.09GB peak 21.51 tok/s with native MTP speculative decoding—63% faster than no drafter in my test. The video uses real timestamped stream chunks. Exact Hugging Face recipe ↓
Qwen 3.8 is a game-changer!
It's performing on par with the latest Opus model. Seriously impressed with the capabilities here.
Finally tested Alibaba_Qwen Qwen 3.8 27B on two RTX 5090 GPUs.
🔥 Getting around 120–130 tokens/sec with vLLM, UnslothAI NVFP4, native MTP, and a 150K context target. The coding output looks very good.
After deploying Qwen 3.8-27B locally, I learned another knowledge
Dgx spark is not suitable for running Dense models Different model architectures have different throughput requirements for underlying computing power and memory bandwidth. If a dense model does not do quantification and KV management well, no matter how powerful the hardware is, it will not work. Sure enough, practice brings true knowledge, let’s feel the speed.
You have reached the end of the archive
All of qwen38