Bring your own llama.cpp fork. No compiling. No Electron. No Python. Point Claude Code at your own machine in one command — fully offline.
Local-LLM tools make two choices for you, and both cost you performance. TurboLLM does the opposite.
Point it at any llama.cpp-compatible binary — a build you compiled, a community fork, or the one it auto-provisions for your GPU. The fastest community innovations land in forks first.
Benchmarks on load, derives fast defaults, and shows a VRAM-fit verdict before you load — no more flag guessing.
Speed in the model list is measured on your machine from actual generation — live while you chat, and remembered per model.
OpenAI and Anthropic-compatible — so Claude Code and every existing tool work unchanged.
No account, no backend, no internet, no telemetry. Your prompts, chats, files, and keys never leave your machine.
The UI runs in the browser, so any phone, tablet, or laptop on your LAN can use the model on your GPU box.
Chat is above. Here's the rest of the app — real screens, doing real work.
Point it at the GGUF folders you already have — LM Studio, Ollama, or a manual download — and it reuses them, no re-downloading. Every model shows a VRAM-fit verdict and its measured tokens/sec, from actual generation on your GPU, so you pick a quant that fits before you commit. One click to Load.
See in docs
Edit any built-in agent's system prompt, skills, or tool access in place — Reset
undoes it — or build your own from a name, a prompt, and a checklist. New in
v1.7.3: a shared Skills library (Claude-style SKILL.md) any
conversation can turn on, plus an MCP servers list your agents can draw tools from.
OpenAI- and Anthropic-compatible endpoints on the same port, API keys for LAN
sharing, and a one-command CLI hookup — turbollm launch claude points
Claude Code at your own GPU with no cloud key.
Practical, step-by-step — using TurboLLM, extending it, and building on top of it.
One command points Claude Code (or opencode, Kilo, openclaw, Hermes) at your local model — no cloud key.
Point TurboLLM at any llama-server-compatible binary — stock llama.cpp, a fork, or your own build.
Upload a SKILL.md, learn one from a folder of docs, or distill one from an existing conversation.
Build your own persona from a name, a system prompt, and a skill/tool allow-list.
OpenAI- and Anthropic-compatible endpoints, API keys, and a model gateway that loads on the fly.
Bind to your LAN and use the model on your GPU box from a phone, tablet, or laptop.
Same GPU (RTX 5070 Ti 16 GB), same model, same 200K context — measured generation speed.
| Qwen3.6-35B-A3B · 200K | TurboLLM | LM Studio | Speed-up |
|---|---|---|---|
| official llama.cpp — q4_0 | 74.7 t/s | 61.0 t/s | 1.2× |
| official llama.cpp — q8_0 | 72.3 t/s | ~66 t/s | 1.1× |
| TurboQuant fork — turbo4 | 24.6 t/s | 11.4 t/s | 2.2× |
Focused on the differences that matter — all four are good tools, and the others move fast.
| TurboLLM | LM Studio | Ollama | Open WebUI | |
|---|---|---|---|---|
| Run any engine / forks | ✓ | ✗ | ✗ | ✗ |
| Benchmark-based auto-tune | ✓ | ◐ | ◐ | ✗ |
| Measured t/s in model list | ✓ | ◐ | ◐ | ✗ |
| Anthropic API → Claude Code | ✓ | ✓ | ✓ | ✗ |
| OpenAI-compatible API | ✓ | ✓ | ✓ | ◐ |
| Lightweight (no Electron / Python) | ✓ | ✗ | ✓ | ✗ |
| Offline-first, no telemetry | ✓ | ◐ | ✓ | ✓ |
No installation, no setup. Just run it.
npx turbollm
Or install globally: npm install -g turbollm
Free and open, built and maintained by one person. If it saves you time, a bit of support keeps it moving.