全网最全的 AI 破甲合集:一个仓库破甲所有模型。收录 1733 个上游破甲项目,9 条可一键安装的破甲卡覆盖 GPT/Claude/Gemini/DeepSeek/GLM/Kimi/Grok/Qwen 与本地模型,内置 llms.txt 让 Claude Code/Codex/Cursor 自己装 有用的话可以点个收藏星星,会持续收录
THE MOST POWERFUL OFFENSIVE AI AGENT FOR PENTESTING , OSINT , AND ETHICAL HACKING
Self-hosted MCP server for image generation and editing with the uncensored Qwen-Image-2.1 model
A proxy tool to bypass GFW.
Runtime behavioral ablation for compressed LLMs. First target: Ternary Bonsai 2 27B — no weight modification or re-quantization. by OrcaRouter team
Qwen3.8-Flash-Next on 2× RTX 3090: with 128 GB RAM up to 4,191 tok/s prefill · 111.5 tok/s decode (131K prompt), full 256K window at 2,865 tok/s; with 64 GB RAM 3,410 tok/s prefill · 84 tok/s decode. New: opt-in uncensored mode (runtime abliteration, no new weights).