AI Agents
openvla
Open-source 7B vision-language-action model and codebase for generalist robot manipulation.
★ 7.1k starsMIT
View guide →Physical Intelligence's open vision-language-action models (pi0, pi0-FAST, pi0.5) for robot policies.
openpi is an open-source ai agents project by Physical-Intelligence (Physical-Intelligence/openpi on GitHub). It has 14,139 GitHub stars and 2,536 forks, is released under the Apache-2.0 license, and was last updated on GitHub on 24 August 2026.
Confirm you have an NVIDIA GPU with enough memory for your goal (inference, LoRA or full fine-tuning) and an Ubuntu 22.04 setup.
Clone the repo including submodules, then use uv to sync dependencies and install the package in editable mode.
Choose a base model for fine-tuning, or an expert checkpoint for your platform such as pi0-FAST-DROID or pi0.5-LIBERO.
Run a trained policy on observations directly or through the policy server, or convert your data to a LeRobot dataset and fine-tune with the provided training scripts.
pi0 (flow-based), pi0-FAST (autoregressive, using the FAST action tokenizer) and pi0.5 with knowledge insulation for better open-world generalization.
Base models pre-trained on 10k+ hours of robot data, ready to fine-tune to your own robot and tasks.
Fine-tuned models for DROID and ALOHA tasks (tabletop manipulation, towel folding, pen uncapping) and a pi0.5 checkpoint for the LIBERO benchmark.
Convert data to LeRobot format, compute normalization statistics, train with the provided configs, and serve the result - including a DROID training pipeline.
Run the model on a separate GPU server and stream actions to the robot over a websocket, keeping robot and policy environments separate.
PyTorch implementations of pi0 and pi0.5 (validated on LIBERO for inference and fine-tuning) alongside the original JAX versions.
Clone openpi - the README asks you to include submodules, so add --recurse-submodules or update them after cloning.
git clone https://github.com/Physical-Intelligence/openpi.git
cd openpiSet up the environment and install the package in editable mode (the LFS flag is needed for the LeRobot dependency).
GIT_LFS_SKIP_SMUDGE=1 uv sync
GIT_LFS_SKIP_SMUDGE=1 uv pip install -e .After choosing or training a checkpoint, launch the policy server to answer observation requests on port 8000.
uv run scripts/serve_policy.py policy:checkpoint --policy.config=pi05_libero --policy.dir=checkpoints/pi05_libero/my_experiment/20000 # example config - see READMEopenpi is published by the Physical Intelligence team, the company behind the pi0 family of robot models.
Yes. The code is Apache-2.0 licensed and the base and expert checkpoints are publicly downloadable.
No. Models run locally on your GPU; checkpoints download from the project's public storage at no cost.
Both are open VLA projects; openpi provides Physical Intelligence's pi0-series models and training/serving pipeline, while OpenVLA is built around its own 7B model and ecosystem - researchers often evaluate both on their platform.
Clone with submodules, run the uv install commands, then follow the example notebook or the no-robot inference test script before connecting real hardware.
Open-source 7B vision-language-action model and codebase for generalist robot manipulation.
Program language models with code and compilers — not fragile prompts.
A blazing-fast enterprise AI gateway: 1000+ models, sub-100µs overhead.
AI-powered reverse engineering assistant that bridges IDA Pro with language models through MCP.
Blazing-fast business intelligence built for humans and AI agents alike.
Turn messy documents into clean, structured data for language models.
The most-starred other AI Agents projects in the LibreAI catalogue — the realistic shortlist if openpi is not the right fit.
A personal AI that really does things — on any OS, with your data staying yours.
An agentic skills framework & software development methodology that works.
Agent harness optimization system with skills, memory and security for Claude Code, Codex and more
The agent that grows with you — a Python framework for adaptive AI agents.
DeepSeek's plugin-based agent harness where everything is a plugin.
Free and open — grab the source, star the repo, and start creating.