Tools
llama.cpp vs LocalAI
llama.cpp or LocalAI? Their plans monthly and yearly, the capabilities their makers state, security and latest updates, side by side — read from the makers' own pages.
In inglese
In short
- Both: Command line, Choice of models, Self-hosted, API
- Only LocalAI states: Runs commands, VS Code, Runs models for you, Builds agents and workflows, Search over your data
llama.cpp
Open-source C/C++ software for running large language and vision-language models locally or in the cloud.
Plans
Free
Free
- Unlimited public/private repositories
- Dependabot security and version updates
- 2,000 CI/CD minutes/month
- 500MB of Packages storage
- Issues & Projects
- Community support
Team
- $4 per user/month
- Access to GitHub Codespaces
- Repository rules
- Multiple reviewers in pull requests
- Draft pull requests
- Code owners
- Required reviewers
Enterprise
- Starting at $21 per user/month
- Data residency
- Enterprise Managed Users
- User provisioning through SCIM
- Enterprise Account to centrally manage multiple organizations
- Environment protection rules
- Repository rules
Prices checked 2026-09-24 on the maker’s page.
Capabilities
- Command line — “llama cli -hf ggml-org/Qwen3.5-0.8B-GGUF” source
- Choice of models — “llama cli -hf ggml-org/Qwen3.5-0.8B-GGUF” source
- Self-hosted — “locally and in the cloud.” source
- API — “Launch OpenAI-compatible API server” source
Latest updates
- b11160
Adds Vulkan int8 cooperative-matrix quantized matmul for AMD RDNA3 and RDNA4, including q8_0, q4_1, q5_0, and q5_1.
- b11159
Handles misalignment in conv_2d and conv_3d.
- b11158
LocalAI
LocalAI runs AI models on your own hardware through an OpenAI-compatible local server and web interface.
Plans
Not read from the maker’s page yet.
Prices checked 2026-09-25 on the maker’s page.
Capabilities
- Runs commands — “It runs shell commands behind an approval gate you control, delegates to sub-agents, and loads MCP servers, plugins and skills.” source
- VS Code — “Install on openSUSE and drive it from VS Code” source
- Command line — “Run local-ai chat and you are talking to an agent that already knows where your models are.” source
- Choice of models — “Every tier of every model we quantize, ranked against the hardware you actually have and installed with one click.” source
- Self-hosted — “keep your data on your hardware, and scale to a room full of GPUs when you need more capacity.” source
- API — “One binary with an OpenAI-compatible API in front of it.” source
- Runs models for you — “Point an existing client at it and the calls keep working, except now the model is on your machine.” source
- Builds agents and workflows — “Run local-ai chat and you are talking to an agent that already knows where your models are.” source
- Search over your data — “Agents, MCP, skills, RAG, interactive tools” source
Latest updates
- v4.10.0 (v4.10.0)
Added a fleet operations dashboard, credentials.yaml authentication, and CLI end-to-end latency and throughput benchmarking.
- v4.9.0 (v4.9.0)
Authentication now defaults to deny; chat supports context compression, canonical model/backend pages, and video serving in vllm-cpp.
- v4.8.0 (v4.8.0)
Added the vllm-cpp backend, 3D generation, a multi-family audio.cpp engine, hardware-matched gallery builds, and distributed-mode fixes.