TurboLLM
Run any local LLM engine, auto-tuned to your GPU — polished web UI + OpenAI/Anthropic-compatible API. Point Claude Code at your own machine in one command. No E
What is TurboLLM?
Run any local LLM engine, auto-tuned to your GPU — polished web UI + OpenAI/Anthropic-compatible API. Point Claude Code at your own machine in one command. No Electron, no Python, offline-first.
Overview of TurboLLM
What it is
TurboLLM is a tool that runs any local LLM engine, auto-tuned to your GPU, with a polished web UI and an OpenAI/Anthropic-compatible API. It is a lightweight ~7 MB npm package on Node, with no Electron, no bundled Chromium, and no Python. It downloads only the engine your GPU needs (Vulkan ≈ 38 MB). It is offline-first and requires no account. The one command npx turbollm starts a local daemon, opens a browser UI, and serves models over an API.
Key features
- Supports any llama-server-compatible engine, including community forks, by probing the binary's real capabilities.
- Auto-tunes launch flags (
-c,-ngl, etc.) to your GPU via benchmarking on load, and shows a VRAM-fit verdict before loading a model. - Provides a web UI and drop-in APIs compatible with OpenAI and Anthropic, so tools like Claude Code work unchanged.
- Can use existing GGUF folders from LM Studio or manual downloads without re-downloading; also supports browsing and downloading from Hugging Face with resume and SHA-256 verification.
- Measures real tokens/sec on your machine during generation and remembers it per model.
- Offline-first and private; telemetry defaults to on but can be changed in Settings.
- Requires Node ≥22 and one terminal command; no packaged installer yet.
Use cases
TurboLLM is for users who are tired of hand-tuning launch flags for every model, want to use community forks (e.g., low-bit KV cache, speculative decoding) without compiling, or want to point agents or Claude Code at a local model with one command. It is also for users who already have folders of GGUFs from LM Studio or Ollama and do not want to re-download them. It is not for users who want a zero-terminal, one-click desktop installer (planned but not yet available) or who need training or fine-tuning (inference only).
Getting started
bash npx turbollm This command requires Node ≥22. It starts a local daemon, opens a browser UI, and serves models over an API.
License and pricing
The license is FSL-1.1 (source-available). It is free for personal and internal business use. Every release converts to Apache-2.0 two years after it ships. The pricing type is open source.
Based on the project's public README and repository metadata.