Tabby

At a glance

Best for

A team or solo setup where completions and chat stay on your own server.

Inference engine

• runs its own models

GPU support

Runs on your own GPU, for example NVIDIA CUDA

Models

Its own registry of code models, such as StarCoder, CodeGemma and Qwen

Features

• Agent workflows
• OpenAI-compatible API

Platforms

Self-hosted server with VS Code, JetBrains and Vim extensions. It runs its own models, so no separate model server is needed.

Sources: tabby.tabbyml.com

Install

docker run -it -p 8080:8080 -v $HOME/.tabby:/data tabbyml/tabby serve --model StarCoder-1B --chat-model Qwen2-1.5B-Instruct

docker run -it --gpus all -p 8080:8080 -v $HOME/.tabby:/data tabbyml/tabby serve --model StarCoder-1B --device cuda --chat-model Qwen2-1.5B-Instruct

Tabby serves its own models, so you do not need Ollama or llama.cpp. Pick the model with --model and open http://localhost:8080 to finish setup.

Alternatives

Pi Hermes Agent OpenClaw OpenCode Codex CLI OpenHands Zed Cline Open Interpreter Goose Aider Crush Qwen Code Kilo Code Open WebUI AnythingLLM Khoj Letta

Which app should I use?