gem-llm
A Gym for Generalist LLMs.
Decision gist · record as of 2026-08-14
Yes, with conditions. GEM is worth installing if you are actively researching or building agentic LLM training systems and need a standardized, framework-agnostic environment API. The low install friction, permissive Apache-2.0 license, and diverse task collection make it a solid foundation. However, the aging maintenance status (last commit 2026-01-21, no recent releases) and beta stability warrant caution for production use. Verify that the runtime dependencies and supported RL frameworks remain compatible with your workflow before committing to it.AI-flagged interpretation of the facts on this page — verify before relying
Before you install
- Requires Python 3.10, 3.11, or 3.12 (capped at 3.12).
- The package depends on nltk, math-verify, reasoning-gym, and msgspec; ensure these are installed or let pip resolve them.
- Low installation friction with a pure-Python wheel.
License · maintenance · safety
Apache-2.0 (permissive) — Licensed under Apache-2.0 (permissive), allowing commercial and private use with minimal restrictions. You may use, modify, and distribute the package freely provided you include a copy of the license and state significant changes.
last release 2025-10-05 (313 days) · last repo commit 2026-01-21 · 506 stars
0 known vulnerabilities (OSV.dev, 2026-08-14) · 224,504 downloads/mo, #9,232 on PyPI
Alternatives
Verify before relying
pip install gem-llm
import gem
env = gem.make("game:GuessTheNumber-v0")
observation, info = env.reset()
action = env.sample_random_action()
next_observation, reward, terminated, truncated, info = env.step(action)- Whether the aging maintenance status (last commit 2026-01-21, no recent releases) indicates the project is still actively maintained or has entered maintenance-only mode.
- Compatibility and stability of the four runtime dependencies (nltk, math-verify, reasoning-gym, msgspec) with the current gem-llm version.
- Whether the supported RL training frameworks (Oat, Tinker, Verl, RL2, ROLL, OpenRLHF) are all actively maintained and compatible with gem-llm 0.1.0.
What it is and what it does
GEM is an open-source environment suite designed for training agentic LLMs via online reinforcement learning. It provides a standardized API modeled after OpenAI Gym, offering a growing collection of diverse environments across games, mathematical reasoning, competitive coding, knowledge-intensive QA, and synthetic reasoning tasks. The package supports tool integration (Python execution, search, MCP), async vectorized execution for high-throughput simulation, and is training framework-agnostic, with demonstrated integration examples for supported RL frameworks.
The core use case is enabling researchers and practitioners to train LLM-based agents through interactive experience rather than static datasets. GEM handles environment management, observation-action-reward loops, and episode termination logic, while remaining agnostic to the choice of training algorithm (REINFORCE, GRPO, PPO, REINFORCE + ReBN) or RL framework. It depends on nltk, math-verify, reasoning-gym, and msgspec for core functionality.
Use it for
- Train agentic LLMs on game-solving tasks like Sudoku or number-guessing using GEM's built-in game environments and your choice of RL framework.
- Develop and benchmark LLM agents on mathematical reasoning problems by using the math environment category and comparing algorithm performance.
- Build tool-using agents that call Python executors or search APIs within GEM's environment wrapper system for complex multi-step tasks.
- Integrate GEM environments with existing RL training frameworks to leverage your preferred infrastructure for distributed training.
- Evaluate LLM agent performance across diverse reasoning tasks from the ReasoningGym category without reimplementing evaluation logic.
Worth the install?
AI-flagged interpretation of the facts on this page. Verify before relying on it.
Yes, with conditions.
GEM is worth installing if you are actively researching or building agentic LLM training systems and need a standardized, framework-agnostic environment API. The low install friction, permissive Apache-2.0 license, and diverse task collection make it a solid foundation. However, the aging maintenance status (last commit 2026-01-21, no recent releases) and beta stability warrant caution for production use. Verify that the runtime dependencies and supported RL frameworks remain compatible with your workflow before committing to it.
Install
gem-llm on PyPI
Before you install
Low installation friction with a pure-Python wheel. The package is in beta (Development Status 4) and aging—last commit was 2026-01-21. The four runtime dependencies (nltk, math-verify, reasoning-gym, msgspec) are lightweight, but the lack of recent updates may signal limited active support.
Requires Python 3.10, 3.11, or 3.12 (capped at 3.12). The package depends on nltk, math-verify, reasoning-gym, and msgspec; ensure these are installed or let pip resolve them.
License in practice
Licensed under Apache-2.0 (permissive), allowing commercial and private use with minimal restrictions. You may use, modify, and distribute the package freely provided you include a copy of the license and state significant changes.
Quickstart
pip install gem-llm
import gem
env = gem.make("game:GuessTheNumber-v0")
observation, info = env.reset()
action = env.sample_random_action()
next_observation, reward, terminated, truncated, info = env.step(action)Verify before relying
- Whether the aging maintenance status (last commit 2026-01-21, no recent releases) indicates the project is still actively maintained or has entered maintenance-only mode.
- Compatibility and stability of the four runtime dependencies (nltk, math-verify, reasoning-gym, msgspec) with the current gem-llm version.
- Whether the supported RL training frameworks (Oat, Tinker, Verl, RL2, ROLL, OpenRLHF) are all actively maintained and compatible with gem-llm 0.1.0.
Package facts
| License | Apache-2.0 permissive |
| Python support | Capped below the current Python release <=3.12,>=3.10 |
| Install friction | Low. Pure-Python wheel |
| Runtime dependencies | 4 packagesnltkmath-verifyreasoning-gymmsgspec |
| Maintenance | Aging 313 days since the last release |
| Last repo commit | |
| First released | |
| Downloads | 224,504 / month, #9,232 on PyPI 30-day window, as of 2026-08-14 |
| Known vulnerabilities | None known OSV.dev, checked 2026-08-14 |
| Classifiers | Development Status :: 4 - BetaProgramming Language :: PythonProgramming Language :: Python :: 3.10Programming Language :: Python :: 3.11Programming Language :: Python :: 3.12Programming Language :: Python :: Implementation :: CPythonProgramming Language :: Python :: Implementation :: PyPy |
Evidence: gem_llm-0.1.0-py3-none-any.whl
Tags
Let your AI agent find packages like this
Example. Real query, live index.
You found this page by searching. An agent finds it by wishing: SkillFed indexes 14,416 PyPI packages by what they can do, searchable in plain language.
wish › “llm reinforcement learning environment”
- gem-llmGEM is a reinforcement learning environment suite that provides a…
- verifiersVerifiers provides environments and evaluation harnesses for training…
- nemo-gymNeMo Gym provides infrastructure for building, running, and scaling…
Give your agent the search over MCP, or paste the wish link into any chat.
More Artificial Intelligence packages
LiteLLM provides a unified Python interface to call 100+ LLM providers (OpenAI, Anthropic, Gemini, Bedrock, Azure, and others) using OpenAI-compatible API format, available as both a Python SDK and a self-hosted AI Gateway proxy server.
Install it if you need to work with multiple LLM providers or want to centralize LLM routing in your organization.
Client library and CLI tool for downloading, uploading, and managing models, datasets, and repositories on the Hugging Face Hub platform.
Install it if you work with Hugging Face Hub models or datasets.
LangChain provides a framework for building agents and LLM-powered applications by composing language models, tools, and memory through a unified API that abstracts over multiple model providers.
hf-xet provides chunk-based deduplication and efficient file transfer for the Hugging Face Hub, enabling faster uploads and downloads of large files with local disk caching.
Tokenizers converts raw text into token sequences for NLP models, with support for training custom vocabularies and using pre-built tokenizers (BPE, WordPiece) optimized for speed via Rust.
Transformers provides a unified framework for loading, fine-tuning, and running state-of-the-art pretrained models across text, vision, audio, video, and multimodal tasks using PyTorch, JAX, or TensorFlow.
Install it if you need to run or train any transformer-based model for NLP, vision, audio, or multimodal tasks.
See also TextArena · verl · reasoning-gym · nemo-gym · kaggle-environments · verifiers · fhaviary · tinker_cookbook · Shimmy · gymnasium