AnalysisAI ModelsOctober 15, 2026

Sabi-Noema fine-tunes Qwen 3.8 27B for capability and efficiency

Read original source →reddit.com

Two post-trained versions of Qwen 3.8 27B, released as Sabi-Noema, target both small-model capability and reasoning efficiency. Weights are published on Hugging Face under sabicap/sabi-noema.

1 source

More stories today

Open the live feed