Sabi-Noema: open Qwen 27B fine-tune cuts tokens 50% and beats bigger models on agentic benchmarks
ExperienceDry6608 · reddit · 2026-10-15
The Sabi-Noema team released two open post-trains of Qwen 3.8 27B focused on small-model capability and reasoning efficiency, trained exclusively on open-source environments. Sabi-Noema beats GPT 6 Luna-medium on DeepSWE and Luna-high on TB2.1, Tau-3, ClawEval and BFCL v4, with a notably lower hallucination rate. It generates 50% fewer tokens per response on TB2.1 and 25% fewer on DeepSWE. The -MLA variant applies a training-free post-hoc MLA conversion of Qwen's GQA attention: 3.6x fewer bytes per token at BF16, and 8.4x smaller KV cache at FP8 — enabling 534K context on a single 24GB GPU at q40 vs 150K for base Qwen. Both models are on Hugging Face.
More from Models
- Opus 5.5 still leads a month later, but open-weight decision models are closing in — airesearch12 · 2026-10-16
- Google restricts Gemini Pro access for Google Plus subscribers, matching free tier — EmergencyLess8861 · 2026-10-16
- Meta releases Muse Spark Open, trained with vLLM-powered RL inference from day one — colinmcnamara · 2026-10-16
- Meta open-sources Muse Spark: runs on two Mac Studios, license bans passing output off as human-written — mawcopolow · 2026-10-16
- Golden Gate Claude explainer resurfaces: interpretability premise still holds — HarperSCarroll · 2026-10-16
- Hugging Face livestreams release of a new open-weight model — huggingface · 2026-10-16