Hands-on with Qwen3.6-35B-A3B: small experts make it blazing fast locally
OwariDa · x · 2026-10-02
A user shares hands-on impressions: qwen3.6-35b-a3b isn't the smartest in its lineup, but its 3B-sized experts make it remarkably fast. It works well for plenty of tasks, especially when you want to guarantee your data never touches any cloud provider.
More from Models
- Claude Skills, PI Code Mode, Codex plugins, DeepSeek harness: super agents converge on plugin systems — op7418 · 2026-10-02
- User burns 14+ hours of Codex usage after GPT suddenly stops following basic instructions — Airaneon · 2026-10-02
- ThursdAI: GPT-6.1 Sol, Sonnet 5.5 and Gemini 4 Argon all land as nobody paces the frontier — altryne · 2026-10-02
- "Opus 5.5 never made mistakes at launch" is nonsense, argues Redditor pushing back on nerf claims — japt77 · 2026-10-02
- Claude Max Feels Basically Unlimited on Opus and Sonnet, Says Peter Yang — petergyang · 2026-10-02
- Quota resets arrive 10AM PST as users rush to burn remaining Ultra limits — Bloated_Plaid · 2026-10-02