OpenJarvis
#6 in Lokale LLM-Runtimesollama · siet 2026-03-12 · 5× · tolest 29. Juni 2026
25
Momentum
OpenJarvis is an open-source framework for building local AI agents, developed by Stanford's Hazy Research and Scaling Intelligence Lab as part of their "Intelligence Per Watt" research. Inference runs on-device by default; cloud APIs are optional and called only when necessary. The framework is organized around five swappable primitives (Intelligence, Engine, Agents, Tools & Memory, Learning) and supports multiple inference backends as well as a continuous local learning loop. Version 1.0 was released on March 12, 2026 under Apache 2.0.
Momentum-Verloop
04.04.03.07.
Features
| API Type | CLI (jarvis ask), Python SDK, FastAPI server (OpenAI-compatible drop-in replacement) |
| Inference Backend | Local: Ollama, vLLM, SGLang, llama.cpp; Cloud (optional): OpenAI, Anthropic, Google Gemini, OpenRouter, MiniMax – all via a unified engine interface |
| Maximum Model Size (GB RAM) | Minimum 8 GB RAM, 16 GB RAM recommended; optional GPU acceleration via NVIDIA RTX 3060+ or Apple M1+ |
| Platforms (OS Support) | macOS, Windows (native + WSL2), Linux |
| Price Tier | Free, open source (Apache 2.0) |
| UI Type | CLI, browser app (localhost:5173), native desktop app (.dmg / .exe / .deb / .rpm / .AppImage) |