Local inference
Runs language models on your own laptop or servers instead of sending data to a hosted provider.
Search AirStackWhy teams use it
- Data stays on machines you control
- Work offline or in restricted networks
- No per-request charges once you own the hardware
- Try open models before you commit to one
Who uses it: ML engineer, AI engineer, Platform engineer, Software engineer, Data scientist, Researcher. Watch for: Smaller local models may not match hosted quality, and serving many users needs real GPUs; test on your own tasks first. AirStack’s guide to the category.
11 products
| Product | What it says it does | Starting at |
|---|---|---|
| llamafile | Distribute and run LLMs with a single file. | Not observed |
| LM Studio | The agent for your computer | Not observed |
| LocalAI | Make AI run on every machine. | Not observed |
| Jan | Your personal intelligence. Run any model, your way. | Not observed |
| llama-cpp-python | Python bindings for llama.cpp | Not observed |
| vLLM | The High-Throughput and Memory-Efficient inference and serving engine for LLMs | Not observed |
| SGLang | High-performance serving framework for large language and multimodal models. | Not observed |
| MLX LM | Run LLMs with MLX | Not observed |
| Ollama | Use open models in your desktop apps and coding agents, or build them into your application. | Not observed |
| MLX | MLX is a NumPy-like array framework designed for efficient and flexible machine | Not observed |
| llama.cpp | LLM inference in C/C++ | Not observed |
Prices are list prices as written on each pricing page; hover one for its quote. “Not observed” means AirStack has not found it on that product’s pages.
Open models for the job
- openai-community/gpt2 30,091,264 downloads in 30 days · mit
- Qwen/Qwen3-0.6B 22,322,522 downloads in 30 days · apache-2.0
- Qwen/Qwen3-14B 12,949,368 downloads in 30 days · apache-2.0