Qwen 35B Fine-Tune Shows Improvements
habachilles · reddit · 2026-07-13
The author is working on a Qwen 35B fine-tune project, aiming to train on as much "coherent" fable data as possible. Current progress includes: - Renting several H200s for training; - Being highly selective with the data to ensure coherence; - Observing around a +5% improvement on both human eval and SWE benchmarks so far; - Planning to open-source the model and publish metrics after a few more runs. The author also asks the community to contribute fable data, believing these datasets are "already sitting on everyone's hard drives" and can help make local LLMs stronger.
More from Models
- Claude 20x users report sharply tighter limits and faster quota burn — MarcJSchmidt · 2026-07-21
- Cola launches July, the latest model jokingly billed as “second only to Fable” — oran_ge · 2026-07-21
- Kimi K3 looks stronger and about 5× cheaper on a frontend dashboard task — OwariDa · 2026-07-21
- Last Week in AI recap: Anthropic’s $65B round, IPO filing, and Microsoft’s MAI push — Last Week in AI · 2026-07-21
- A user says Claude 4.6 felt worse yesterday and asks whether model quality can drift over time — Rahios · 2026-07-21
- Kimi K3 hits 89.4% peak on software tasks while Fable 5 is slightly steadier — FinanceYF5 · 2026-07-21