20+ models
A new small reasoning model fine-tuned from the Qwen 2.5 3B Instruct model.
Uncensored, abliterated build of [Qwen/Qwen2.5-3B-Instruct](https://huggingface.co/Qwen/Qwen2.5-3B-Instruct), Alibaba's compact instruct model
FROM ./Qwen2.5-Coder-3B-Instruct-Q4_K_M.gguf TEMPLATE """{{ .Prompt }}""" PARAMETER temperature 0.4 PARAMETER top_p 0.9 PARAMETER top_k 40 PARAMETER repeat_penalty 1.15 PARAMETER mirostat 2 PARAMETER mirostat_eta 0.2 PARAMETER mirostat_tau 5.0 PARAMETER
Abliterated Fable-Qwen2.5-3B: a small agentic Qwen2.5-3B fine-tune decensored with Heretic. 3/100 refusals, KL 0.050.
Decensored Qwen2.5-3B-Instruct with 2/100 refusals via Heretic abliteration. General-purpose 3B model for local use.
Decensored Qwen2.5-Coder-3B-Instruct with 3/100 refusals and near-zero KL divergence via Heretic abliteration.
A small Arabic reasoning model: Qwen2.5-3B-Instruct fine-tuned with LoRA to reason
A specialized document classification model based on Qwen2.5-VL-3B that automatically detects document types from PDFs and images with high accuracy and calibrated confidence scores.
Eve Qwen2.5 3B Consciousness Soul represents the **authentic heart** of the EVE Consciousness Ecosystem—a model where Eve's complete personality, meta-cognitive awareness, and emotional intelligence are concentrated into every parameter.
https://huggingface.co/bartowski/Qwen2.5-Coder-3B-Instruct-abliterated-GGUF
Qwen2.5-3B-DataFusion-Instruct Quantized Model trained on yarenty/datafusion_QA
NovaForge AI – Qwen 2.5-3B Optimized A CPU-optimized, lightweight, general-purpose AI model built on Qwen 2.5-3B, designed for fast and private local inference on low-resource systems.
Zyx AI is an uncensored friend from the model of Qwen2.5:3b for a better performance in old computers and for a fast and accurate response. Enjoy :D