Thứ Tư, 19 thg 8, 2026

Nhờ áp dụng kỹ thuật 1-Bit Quantization, mô hình Qwen3.8-27B có thể vận hành trên các thiết bị chỉ có 8GB RAM nhưng vẫn duy trì được 77% độ chính xác so với bản gốc.

Đăng nhập để góp ý, chỉnh sửa
Nguồn chính
SOURCE@danielhanchen15:51 19 thg 8Divergence-300 is a new metric which extends top-1% greedy acc to 32 tokens & more, and we used 300 unseen examples from Terminal Bench, DeepSWE and more
SUPPORT@danielhanchen16:50 19 thg 8We originally didn't want to release them but were shocked that they worked extremely well on our internal testing!
SUPPORT@tinkerapi17:29 19 thg 8It’s natively multimodal with images + video, has flexible thinking control, and is meaningfully better at coding, professional work, research, and long-horizon agentic tasks.
SOURCE@alibaba_qwen3:47 19 thg 8Strong enough to handle professional tasks. Small enough to run on your local machine.
SUPPORT@alibaba_qwen3:35 19 thg 84 days to the top.🏆Thank you to every builder who pushed Qwen3.8-27B to #1 on Cline!