RLinf

RLinf is an open-source reinforcement learning infrastructure for embodied and agentic AI, meant for scaling RL training across many GPUs and nodes.

Licenses

Code: Apache-2.0 (opens in a new tab)
Read by Robotheca from LICENSE (opens in a new tab) (Apache-2.0) at commit 6b7853b (opens in a new tab), on 16 Oct 2026.
Commercial use: Yes

Supports diverse RL training workflows such as PPO, GRPO and SAC without requiring changes to user code to scale out.
Offers multiple backend integrations, pairing FSDP with HuggingFace, SGLang or vLLM for rapid prototyping, or Megatron with SGLang or vLLM for large-scale training.
Covers supervised fine-tuning, simulation RL and real-world RL for vision-language-action and world action models.
Lists integrations with many simulators, including ManiSkill, LIBERO, RoboTwin, BEHAVIOR, IsaacLab and CALVIN.
Supports many VLA models, such as π₀ / π₀.₅, OpenVLA-OFT, GR00T and StarVLA, plus VLM and world model families.
Includes an agentic reinforcement learning path with examples for math reasoning and search-based tasks.

Written by Robotheca from the project's README (opens in a new tab) at commit 6b7853b.

Checked on GitHub on 16 Oct 2026.

No robot on Robotheca is linked to it yet. Links come from a robot's robot.yaml, from its dependency files read as text, or from Robotheca's editors.

Robotheca lists software as a reference: nothing is hosted or run here, and every fact links to where it was read.