An open-source implementaion for fine-tuning Qwen-VL series by Alibaba Cloud.

RAG (Retrieval-augmented generation) ChatBot that provides answers based on contextual information extracted from a collection of Markdown files.

An open-source Cotypist with macOS system wide AI autocomplete

Run generative AI models in sophgo BM1684X/BM1688

Qwen3.5-122B-A10B on DGX Spark: 28.3 → 51 tok/s (+80%)

qwen proxy . qwen code oauth file in use for a proxy server so you can use qwen coder model any tool that uses openai api and base url . and also the websearch api proxy

Private on-device AI chat for Android — runs any GGUF model locally via llama.cpp with ARM-optimised SIMD. Zero network permissions, encrypted settings, biometric lock, tamper detection.

Make Local AI Toys, Robots, Devices that work with a MacBook and an Arduino ESP32

MCore-Bridge: Providing Megatron-Core model definitions for state-of-the-art large models and making Megatron training as simple as Transformers — with support for 300+ large language models (Qwen3-Next, GLM-5.1, Deepseek-V4, MiniMax-2.7, ...) and 200+ multimodal large models (Qwen3.5, Qwen3-Omni, Gemma4, ...).

4-5x faster Qwen3.5 on ASUS GX10 / DGX Spark — Hybrid INT4+FP8 + MTP via one shell script

Fine-tune a local LLM on your own app's data in 15 minutes. Runs entirely on-device, zero cloud after training. Apple Silicon + CUDA.

A professional automation toolkit for ComfyUI to prepare LoRA training data using LLMs. New support for Qwen 3.5

Fast Qwen 3.5 inference in pure Java

7.67× LoRA / 8.35× Full FT speedup for Qwen3.5 (0.8B–27B) on NVIDIA DGX Spark — wall-clock parity with rented H100. Lossless within BF16. Three-command interactive wizard handles model picker, data validator, training, and merge.

PulseCoreLite is a desktop performance monitoring application based on Tauri 2 and Vue 3. It offers two forms of monitoring: floating window monitoring and taskbar monitoring. It displays real-time indicators such as CPU, GPU, memory, disk, and network, and supports system-level capabilities such as multiple languages, transparency and refresh rate

SAM3-Plus-Qwen3.5 is an advanced, experimental computer vision suite that seamlessly integrates Facebook's Segment Anything Model 3 (SAM3) with the Qwen3.5 multimodal reasoning engine.

A local-first, privacy-focused CLI that brings vision and semantic understanding to the Unix pipe. Powered by Qwen 3.5-0.8B and Rust. 🦀

Adaptive routing system combining YOLOv8s and Qwen3.5-0.8B VLM for semantic disambiguation of low-confidence detections on VOC2012.

Qwen 3.5 Reverse Proxy for handling instant / thinking modes and their variants automatically

Autonomous agent for BitGN PAC1 Challenge. Schema-Guided Reasoning pipeline on Qwen3.5-27B. 67.3% (70/104) on PAC1-PROD. Includes full competition run logs.