Qwen3.6 35B A3B: Claude 4.6 Opus Reasoning Distilled GGUF
New Qwen3.6 35B A3B distilled on Claude 4.6 Opus data. GGUF quantizations Q4_K_M to Q8_0—aimed at better logic and fewer loops.
Qwen3.6 35B A3B Claude-4.6 Opus Reasoning-Distilled-GGUF
The new qwen3.6, trained and distilled on Opus 4.6 data; I posted something similar for gemma4 — if you’re interested, grab it
Just so you know, this doesn’t mean it writes code significantly better, but at the same time it thinks closer to Opus and gets rid of looping and brain farts
Distillation in general is needed to make the model think, you could say, according to the template of the reference — in our case Opus 4.6; it will be better at logic, I think, and a bit at coding too
Versions Q4_K_M — Q8_0
Comments
Weekly digest
The best of vibe coding, AI agents and open source — once a week, no spam.
Related articles
All articlesQwythos-9B-Claude-Mythos-5 Released with 1M Token Context Window
Empero unveils Qwythos-9B, a Qwen3.5-9B-based reasoning model fine-tuned on 500M+ tokens of Claude Mythos/Fable data and supporting up to 1M context.
Qwen3.6 35B A3B vs Gemma4 26B A4B: MoE Vibe-Coding Face-Off
Two MoE models take the same vibe-coding challenge: Qwen3.6 35B A3B vs Gemma4 26B A4B on Unsloth Q6_K_XL + llama.cpp across 4 prompts.
Alibaba Qwen3.6-35B-A3B: Sparse MoE Model with 3B Active Params
Alibaba’s Qwen3.6-35B-A3B is a sparse MoE with 3B active params, strong coding and multimodal vision results, plus GGUF for local 4-bit runs.