chinese-llm

AirLLM 70B inference with single 4GB GPU

A trilingual (繁中 / English / 简中) learning roadmap for agentic AI: from LLM basics to multi-agent systems, with 240+ curated resources and hands-on examples. 中文 AI agent 學習地圖。

Codex/Claude Code/Cursor的开源增强版，专注一句话实现复杂长程任务（教育、编程、办公、生活、娱乐、游戏）。部署在你自己的服务器上，团队用浏览器打开就能编程——包括手机。CLI & Web UI 双入口，Multi-Agent 协作，原生直连千问/DeepSeek 等国产大模型。技能/插件/跨会话记忆，8 层安全沙箱，数据不离开你的机器。Docker 一键部署，MIT 开源，零锁定。

Python toolkit for Chinese LLMs, with flexible batch capacity, structured real-time visulization and automated accumulation for streaming, and explicit feedback on vendor-native parameter validation.

A research fork of opencode demonstrating Language Anchoring — making LLMs think consistently in your language. Verified: 95%+ Chinese thinking compliance.

Open & nano reimplementation of Sakana Fugu. A repo-native, multi-agent coding loop powered by 9+ LLMs (isolated via Claude Code) and an independent Codex reviewer. Lightweight, bounded, and self-improving (Self-Harness)—no coordinator training required.

🔌 让 OpenAI Codex CLI 接入 GLM (智谱 AI) | Local proxy enabling Codex CLI to work with GLM models - 流式响应 & 工具调用 / Streaming & Tool Calling

中文公司调研 Agent

re!think it. System prompt teaching LLMs to execute two core tasks: complex answers without hallucinations, and creative ideas without clichés. Written in math-like logic, which LLMs parse better than plain language. Built for mid-to-high complexity tasks, featuring a Bypass branch to execute simple prompts directly without added cognitive overhead

This repo introduces MagicData-CLAM, a Chinese SFT dataset, and provides to the community two relevant models that we finetuned. Contact business@magicdatatech.com for more information.

MoziAI-35B: Local open-source financial AI LLM. 35B MoE model compressed to 15.5GB via MoziSmartBit quantization. 256K context, vision, tool calling, 140+ tok/s on consumer GPU. Based on Qwen3.6-35B-A3B.

🦞 Curated OpenClaw config templates for Chinese LLM providers, multi-channel setups, automation & more | OpenClaw 配置模板大全

基于户晨风直播语料微调的 AI 对话模型

网络安全领域大模型:Qwen3.8-27B LoRA 注入中文漏洞库知识,专家盲审 8.08/10(B 类 +83%)· vLLM+FP8 生产化 · 28 坑踩坑实录

Hands-on local LLM fine-tuning course (SFT / LoRA / PEFT / DPO / QLoRA) with a browser studio that visualizes loss, tensors, and adapter diffs.

🎯 Fine-tuning LLMs using LlamaFactory for financial intent understanding | Evaluating open-source models on OpenFinData benchmark | Full implementation with multiple models (Qwen2.5/ChatGLM3/Baichuan2/Llama3)

Chinese Reasoning Language Model with Step-by-Step trajectories.

GhostAI 知识库 — 5 分钟拥有专属 AI 知识库 · DeepSeek + bge-m3 · 国内合规部署 · ¥39 起

0.1B 中文 Decoder-only LLM 端到端训练项目，覆盖 Tokenizer、Pretraining、SFT、GRPO、Benchmark 与 Inference，基于 PyTorch 从零实现模型、数据处理和训练评测流程。

✨ XingLing (星灵): A lightweight 0.68B Chinese Chat LLM built from scratch (Pretraining + SFT)