Models

A Chinese writing model fine-tuned from Qwen3-4B with 4.3 GB weights, 256K context. The model was trained on an in-house Chinese corpus, with a strength in literary criticism, personal essays, and commentary on social phenomena.

OpenAI GPT-OSS 20B Uncensored MoE (20.9B total, 3.6B active, Heretic NEO CODEPlus IQ4_NL, 11.8GB). 320.2 tok/s Decode, 19920 tok/s Prefill, 57.1% Precision, 0.0% Hallucination (6/6), 100% Zero-Refusal. License: Apache-2.0.

Fine-tuned Qwen 3.8 27B (OBLITERATUS/Qwen3.8-27B-OBLITERATED) specialized for offensive and defensive cybersecurity, penetration testing, reverse engineering, and code execution. Trained on 2,000 high-context technical samples with native 256K context, to

Totem is the Total.js assistant: Total.js 5 and jComponent 20, in your language, with short working code. Inside the Totem application it builds complete applications.

Open 4.5B multimodal decision model by HAL-X: send a message, document or image with your options and get a calibrated probability for each one. #2 on Decision Index, 12K+ downloads HF, 20+ languages. Apache-2.0.

Deterministic code generation engine for Syntax Sanctuary (Linux, Docker, Bash, monolithic PHP & Vanilla JS). Upgraded to 7B with 4096 context window. / Motor de código determinista actualizado a 7B.

Microsoft AesCode 32B VL Uncensored Multimodal UI-to-Code Specialist (153-Tensor GGUF Writer Transplant + mmproj-Q8_0 Vision). 64.4 tok/s Decode, 5,680 tok/s Prefill, 85.7% Precision, 0.0% Hallucination (6/6). License: Apache-2.0.

Qwen3-Coder 30B-A3B Uncensored MoE (30.5B total, 3.0B active, Q4_K_M + GGUF <think> patch). 272.5 tok/s Decode, 10,861 tok/s Prefill, 92.9% Precision, 0.0% Hallucination (6/6), 100% Zero-Refusal. License: Apache-2.0.

A work model for inteliigent thinking

594B Mixture-of-Experts model (17B active) with image input and up to 1M tokens of context, built for agentic coding and knowledge work.

Experimental INT4 tool-selection decision model. Uses decision-token scoring; tested on Apple Silicon with Ollama.

self-build model