llama.cpp
By ggml.ai (part of Hugging Face) / open source · github.com/ggml-org/llama.cpp
3,5 Good
The engine that has driven the entire local AI wave: extremely optimised C++ code that runs language models on almost any hardware. For those who build themselves.
Best for: Developers who want maximum control and performance.
Pros
- Fastest and most hardware-flexible (CPU, GPU, Mac)
- Open source under MIT licence, created by Georgi Gerganov (ggml.ai, since 2026 part of Hugging Face)
- The base for Ollama, LM Studio and most others
Cons
- Pure developer product: compile and configure yourself
- No support beyond the community
Quick facts
| Price | Free |
|---|---|
| Origin | Outside EU (ggml.ai (part of Hugging Face) / open source) |
| Categories | Local AI |
| EU data storage | No |
| GDPR terms (DPA) | No |
Sources & evidence
Claims about data storage, jurisdiction and compliance should be verifiable. Here is the evidence behind the statements above:
- ggml-org – llama.cpp on GitHub (open source under MIT, local LLM inference in C/C++ on CPU, GPU and Mac) →
- ggml.ai – Homepage (the company behind ggml, founded 2023 by Georgi Gerganov, acquired by American Hugging Face in 2026) →