NeMo Gym
NVIDIA Apache-2.0 library for building, evaluating and scaling agent environments, verifiers and harnesses, including sandboxed code-execution tasks.
Tags
Description
Established NVIDIA project (repo NVIDIA-NeMo/Gym; not archived; created 2025-08-25; not presented as a brand-new 2026 launch). Root LICENSE is Apache-2.0. PyPI nemo-gym 0.7.0 verified 2026-11-12 (not yanked; wheel uploaded 2026-10-12; project homepage is this GitHub repo). The official README quick start, however, installs from a git clone with uv venv (Python 3.13.14+) and uv sync, not a pip install one-liner; both facts are recorded. Overview docs at https://docs.nvidia.com/nemo/gym/main/about/ (HTTP 200). The get-started path https://docs.nvidia.com/nemo/gym/main/get-started responds 307 to /nemo/gym/main/get-started/prerequisites; this entry uses the final URL https://docs.nvidia.com/nemo/gym/main/get-started/prerequisites (HTTP 200, no further redirect). README describes environments as a dataset, agent harness, verifier and per-task state, including code execution, tool calling and sandboxes, and optional OpenTelemetry tracing. Quick-start examples use a model API key (OpenAI is the documented example) — no keys were supplied and no environment servers were started. GPU is not required for the library itself; specific resource servers or inference may need one. Vendor: NVIDIA. Related training stack: nemo-rl.
Agent access
Related
- NeMo RL: NVIDIA Apache-2.0 post-training / reinforcement-learning library for large-scale model training, with official docs and NGC containers.
- OpenEnv: BSD-3-Clause framework for building and serving isolated agentic RL environments over a Gymnasium-style reset/step/state API.
- Harbor: Framework from the Terminal-Bench creators for evaluating and optimizing agents and models in sandboxed environments, run in parallel across cloud sandbox providers.
- verifiers: Prime Intellect’s MIT library for creating LLM reinforcement-learning environments and evals, with a matching PyPI package.
Guides & comparisons
- RL environment libraries compared — NeMo Gym, OpenEnv and Verifiers (Comparison): Compare NeMo Gym, OpenEnv and Verifiers as libraries for writing and running RL environments: authoring model, isolation, documented trainers, install paths, licences and current versions. Distinct from Harbor/AgentEnv/Seahaven and from the agent RL stacks comparison.
Sources
- https://github.com/NVIDIA-NeMo/Gym
- https://github.com/NVIDIA-NeMo/Gym/blob/main/LICENSE
- https://github.com/NVIDIA-NeMo/Gym/blob/main/README.md
- https://pypi.org/project/nemo-gym/
- https://docs.nvidia.com/nemo/gym/main/about/
- https://docs.nvidia.com/nemo/gym/main/get-started/prerequisites
Machine-readable
- JSON
- Markdown
- Source file on GitHub (edit via pull request)