🔥 Someone got a Qwen3.8-27B fine-tune down to ~7.4GiB on GPU (~2.3 bpw). LiveC… · @QwenAIFans · TwKit

@QwenAIFans

🔥 Someone got a Qwen3.8-27B fine-tune down to ~7.4GiB on GPU (~2.3 bpw). LiveCodeBench: 84.3%, just above Qwen3.6-27B BF16 at 83.9%. Caveat: still trails full Qwen3.8 BF16 (90.3%). The 12GB / 200K claim is not fully tested yet. via Hugging Face 👇 https://huggingface.co/jakeatx/qwen3.8-27b-Swift1.5-Uncensored-ATX-AI2

曝光 868 · 评论 1 · 点赞 20 · 书签 11