Qwen 3.6 35B praised for general agentic tasks

A user praised Qwen 3.6 35B (Unsloth UD_Q4_K_XL quant) for consistent general agentic tasks like smart-home control and messaging, reaching ~120-140 t/s generation on Tesla P100s, though it is weak at coding and holding a persona.

• Basalt inference engine launched with up to 2.6x speedup2026-10-10
• Writer says free local LLM can replace ChatGPT and Claude subscriptions2026-10-07
• User calls for pinned model deployment guides on LocalLLaMA2026-09-27
• Swift1.5-Qwen3.8-Flash-Next shines in benchmarks2026-09-25