DeepSeek-V4-Pro-Qwen3.5-9B 'Heretic': a decensored 9B reasoning model
A community 9B reasoning model distilled from DeepSeek-V4-Pro into Qwen3.5-9B on ~250K math and STEM samples, then decensored with the Heretic ablation method; published on Hugging Face today with GGUF quants already available.
Why it matters
A 9B model claiming 94.5% on GSM8K and 90.5% on MMLU-Pro math/physics/chemistry subsets, distilled from a 1.6T teacher and stripped of refusal layers, small enough to run on a single consumer GPU.
What you could build with it
A researcher could run this 9B reasoning model locally to grade and explain step-by-step maths homework, keeping student work on-device instead of sending it to a hosted API.
Does it hold up?
Too early to judge: published today, with vendor-reported benchmark scores and no independent evaluation yet; GGUF quants already exist, so local runs are possible immediately.
Built with DeepSeek-V4-Pro-Qwen3.5-9B 'Heretic'
- GGUF quants of the Heretic modelhf · Static GGUF quantizations from IQ1_S to Q8_0 plus f16, ready for llama.cpp and Ollama local runs.
- Imatrix GGUF quants of the Heretic modelhf · Importance-matrix GGUF quants for the same model, with a size and quality guide for picking a quantization.
Learn more
First spotted on hf: source.
More AI models
VoiceStudio: open-source, fully-local ElevenLabs alternative for voice cloning, dubbing, and audiobooksdebpalash/VoiceStudio is a desktop app that runs voice cloning, voice design, video dubbing, dictation, transcription…models · JEV 0.75Soup: fine-tune an 8B LLM on a 4 GB laptop GPU from one YAML fileAn open-source CLI that fine-tunes LLMs from a single YAML file, with layer streaming that trains an 8B model on a 4…models · JEV 0.73Tencent Hunyuan Hy3: open 295B MoE with 21B active under Apache 2.0Tencent open-sourced Hy3, a 295B-parameter mixture-of-experts model with only 21B parameters active per token, under…models · JEV 0.73NVIDIA Isaac Lab 3.0 GA + Isaac OS open runtime + Agile ONE SN humanoidMajor Isaac platform expansion at GTC Berlin Oct 21: Isaac Lab 3.0 is generally available as the open-source robot…models · JEV 0.73NVIDIA Nemotron 3.5 Retriever collection: Embed 8B/1B, Rerank 8B, Parse 631MNVIDIA's Oct 21 collection drop for agentic retrieval: Nemotron-3.5-Embed in 8B BF16, 1B BF16 and 1B NVFP4…models · JEV 0.72Kimi K3.1 Pro: Moonshot's 2.8T open-weight flagship officially launchesMoonshot AI officially launched Kimi K3.1 Pro on Oct 23, a mixture-of-experts model with 2.8 trillion total / 104…models · JEV 0.72
Get the week's best AI launches, plus 3 ideas worth building
One email every Saturday. Ranked by traction, not hype. Free.