OpenPipe

license

Apache-2.0(OSI, OpenPipe Inc.

source

public

methods

RL(GRPO)+SFT trajectory training

managed-tier

OpenPipe hosted training/serving SaaS on top of the OSS lib

core-gated

ungated(ART FAQ commits to maintaining ART as a full-featured open source project and calls the hosted ART backend optional, for users who do not want to manage GPU infrastructure

Apache-2.0 with the full source public and nothing withheld from it: no enterprise directory, no closed package the open one depends on, no license key. The ART FAQ commits to "maintaining ART as a full-featured open source project" and calls the hosted backend "an optional hosted ART backend for users who don't want to manage GPU infrastructure on their own", and the docs describe a fully self-hostable LocalBackend beside it. OpenPipe's hosted training and serving SaaS is a separate product sold alongside the open core, which does not count as gating on this scale. Some vendored code carries LGPL-3.0.

• https://github.com/OpenPipe/ART recorded 2026-08-11

  Repo metadata reads `"isPrivate":false`, `"isArchived":false`, `"visibilityLabel":"Public"`, and `"license":{"spdxId":"Apache-2.0","name":"Apache License 2.0"}`.

• https://raw.githubusercontent.com/OpenPipe/ART/main/LICENSE recorded 2026-08-11

  Full Apache License 2.0 text; appendix reads "Copyright 2025, OpenPipe, Inc." followed by an "Additional License Notice" stating the distribution "includes a small portion of code adapted from a third-party project licensed under the GNU Lesser General Public License v3.0."

• https://raw.githubusercontent.com/OpenPipe/ART/main/README.md recorded 2026-08-09

  "Train from anywhere. Run the ART client on your laptop and let the ART server kick off an ephemeral GPU-enabled environment, or run on a local GPU." Install is `pip install openpipe-art`.

• https://art.openpipe.ai/getting-started/faq.md recorded 2026-08-11

  ART FAQ states "We are committed to maintaining ART as a full-featured open source project" and "We've included an open-source ART backend". On the hosted option it says "We will also deploy an optional hosted ART backend for users who don't want to manage GPU infrastructure on their own." The FAQ names no paid OpenPipe plan and no feature paywall.

• https://art.openpipe.ai/fundamentals/art-backend.md recorded 2026-08-09

  Documents two interchangeable backend classes on the same client code: ServerlessBackend "train remotely on autoscaling GPUs" and LocalBackend "run your agent and training code on the same machine" — no capability described as exclusive to either.

openpipe-art averaged 22,537 downloads over the trailing 30 days on pypistats' overall series, inside the 10K-100K band. Its 10,568 GitHub stars corroborate that picture but do not set the level; a measured download figure takes precedence over stars.

• https://pypistats.org/packages/openpipe-art recorded 2026-08-09

  The page's "Overall" Plotly series for openpipe-art covers 2026-02-09 through 2026-08-09; summing its daily values over the last 30 recorded days (through 2026-08-07) gives 22,537 downloads.

A capable, focused agent-RL trainer: multi-step agentic reinforcement learning via GRPO plus SFT on trajectories, running against vLLM and Unsloth backends and supporting Qwen, Llama and GPT-OSS models. It is a specialist rather than general training infrastructure, with no tensor, pipeline or expert parallelism at Megatron-LM's scale, which places it mid-tier among full training stacks. The score rests on that feature comparison; there is no MLPerf Training submission behind it.

• https://github.com/OpenPipe/ART recorded 2026-08-13

  GRPO RL trainer, vLLM/Unsloth integration, Qwen/Llama/GPT-OSS support