prime-rl

---

---

Overview

prime-rl is a framework for large-scale reinforcement learning. It is designed to be easy to use and hackable, yet capable of scaling to 1000+ GPUs. Here is what we think sets it apart:

1. Fully asynchronous RL for high-throughput agentic training at scale.
2. Performant: built to train 1T+ MoE models on 1000+ GPUs with FSDP2 for training and vLLM for inference, with FP8 inference, PD disaggregation, EP and CP parallelism, and more.
3. Native integration with verifiers environments through the Environments Hub, including built-in support for SWE and agentic environments.
4. End-to-end post-training: SFT, RL training, and evals.
5. Multi-node deployment with Slurm support.
6. Multimodal support for Qwen3.5 and Qwen3.8 VLMs, with image RL and SFT examples.
7. Hackable, modular, and extensible by design.
8. One-line SLURM deployment for frontier models — e.g. GLM-5 FP8 with P/D disaggregation, the llm-d router, and Mooncake KV offload.

Models support

The trainer runs the PrimeRL model implementations under src/prime_rl/trainer/models/: dense Llama, Qwen3 and Qwen3.5, plus the families in the table below, with expert parallelism (EP) for MoE layers, context parallelism (CP) for long sequences, and additional kernels like quack-kernels. Other architectures fail at trainer setup.

Family	Example IDs	MoE	EP	CP
GLM-5 (glm_moe_dsa)	zai-org/GLM-5, zai-org/GLM-5-FP8	yes	✅	✅
GLM-5.3 Flash, text only (glm5_next)	zai-org/GLM-5.3-Flash-BF16	yes	✅	✅
Qwen3 MoE (qwen3_moe)	Qwen/Qwen3-30B-A3B, …	yes	✅	✅
Qwen3.5 MoE (qwen3_5_moe)	Qwen/Qwen3.5-35B-A3B, …	yes	✅	✅
Qwen3.5 VLMs	see advanced.md (qwen3_5, qwen3_5_moe)	MoE only on MoE VLMs	MoE only	✅
Qwen3.8 Flash Next VLM (qwen4_exp)	Qwen/Qwen3.8-Flash-Next	yes	✅	✅
Poolside Laguna (laguna)	poolside/Laguna-XS.2	yes	✅	✅
MiniMax M2 (minimax_m2)	MiniMax/MiniMax-M2	yes	✅	✅
Muse Glimmer, text only (muse_glimmer)	meta-models/Muse-Glimmer-30B	no	—	✅
Nemotron H (nemotron_h)	nvidia/Nemotron-3-Nano-30B-A3B, nvidia/Nemotron-3-Super-120B-A12B, …	yes	✅	✅
Trinity (afmoe)	arcee-ai/Trinity-Mini, …	yes	✅	✅
GLM-4 · GLM-4.5 MoE · INTELLECT-3 (glm4_moe)	THUDM/GLM-4-9B-0414, zai-org/GLM-4.5-Air, zai-org/GLM-4.5, PrimeIntellect/INTELLECT-3, …

Setup

We develop and test on NVIDIA RTX 3090/4090/5090, A100, H100, H200, and B200. If your setup fails, please create an issue.

Prerequisites

Currently, you need at least one NVIDIA GPU to use prime-rl. If you don't already have access to one, we recommend our compute platform for everything from renting on-demand single GPUs for developing, debugging and small ablations, to reserving 1000+ GPU clusters for production-scale training.

Quick Setup

Set up prime-rl in a single command.

curl -sSL https://raw.githubusercontent.com/PrimeIntellect-ai/prime-rl/main/scripts/install.sh | bash

1. Clone the repository

git clone https://github.com/PrimeIntellect-ai/prime-rl.git
cd prime-rl

2. Initialize submodules

git submodule update --init -- deps/verifiers deps/renderers deps/prime-envs deps/pydantic-config

3. Install uv

curl -LsSf https://astral.sh/uv/install.sh | sh
source $HOME/.local/bin/env

4. Install dependencies from the lock file

uv sync --all-extras

NOTE: Environments are opt-in uv workspace members — uv sync --all-extras does not install them. To train on them, install all with uv sync --all-extras --all-packages, or a subset with uv sync --package prime-rl --package <env>.

1. Check that the environment uses Python 3.12

uv run python -V

2. Check that flash-attn is installed

uv run python -c "import flash_attn"

3. Check that you can run SFT trainer (this requires 1 GPU)

uv run sft @ configs/debug/fake/sft.toml

4. Check that you can run the inference server (this requires 1 GPU)

uv run inference --vllm.model Qwen/Qwen3-0.6B

5. Check that the full RL stack (inference + orchestrator + trainer) runs end-to-end (this requires 2 GPUs)

uv run rl @ examples/basic/reverse-text/rl.toml

Additional Setup

1. If you want to log your runs to W&B, log in

uv run wandb login
# Or set `export WANDB_API_KEY=...`

2. If you require gated/ private models or datasets from HuggingFace, log in

uv run hf auth login
# Or set `export HF_TOKEN=...`

Training Examples

We provide end-to-end training examples in the examples directory to highlight features of the framework and guide you through the process of training your own models.

Basic Training: 1 to 8 GPUs

Follow this guide to learn the basics of prime-rl. You can train your own models on 1 to 8 GPUs. Ideal for getting started and exploring the capabilities of the framework. These guides cover most use cases -- single-turn, multi-turn, tool calling, etc. -- on toy environments and small models.

1. Reverse Text: Train Qwen3-0.6B to reverse a small chunk of text. Demonstrates tiny-scale single-turn SFT and RL training. Can be trained on a single consumer GPU in a few minutes, and is ideal for getting started.
2. Wordle: Train Qwen3-1.7B to play Wordle. A fun example of multi-turn SFT and RL training. Can be trained on a 2-4 H100 GPUs in a few hours. Ideal for exploring the multi-turn training capabilities of the framework.
3. Alphabet Sort: Train Qwen3-4B-Instruct-2507 to sort names alphabetically. Demonstrates multi-turn RL training via LoRA without SFT warmup. Can be trained on a single H100 GPU in just over an hour. Ideal for exploring LoRA-based training.
4. Wiki Search: Train Qwen3-4B-Instruct-2507 to answer trivia questions by searching through a Wikipedia. Demonstrates multi-turn with web search tool use.
5. Hendrycks Sanity: Run a sanity check experiment on Qwen3-1.7B using a subset of MATH filtered to problems that DeepSeek-R1-Distill-Qwen-1.5B partially solves (20-80%). Useful for algorithm ablations.

Advanced Training: 32 - 2048 GPUs:

Follow this guide to train large models on hard reasoning and agentic tasks. These guides are designed to be run from a Slurm cluster.

1. Qwen3-30B-A3B: Train Qwen3-30B-A3B on math, SWE, and agentic tool use.
2. GLM-4.5-Air: Train GLM-4.5-Air on search, SWE, and terminal tasks.
3. GLM-5: Large-scale GLM-5/GLM-5.3 training with P/D disaggregation, the llm-d router, and FP8.

Docs

Check out the docs directory for in-depth guides on how to use prime-rl.

• Overview - Architecture, install, and a copy-pasteable end-to-end RL run
• Configuration - TOML composition, CLI overrides, env vars, validation
• Training - RL, SFT, checkpointing, observability, rules of thumb
• Eval - Standalone evals against any inference server, online evals, resume, platform upload
• Scaling - Single-GPU through multi-node, FSDP/EP/CP, SLURM, benchmarking
• Algorithms - Async training and staleness, the AIPO loss, advantage and filter plugins, trajectory merging
• Advanced - Custom modeling, multimodal training, LoRA
• Development - Test suite, pre-commit hooks, adding a new model

Contributing

We warmly welcome community contributions! We use issues to track bugs, feature requests, and share our internal roadmap. If you encounter bugs, have pain points during development, or have ideas for new features, please open an issue.

Contributions are welcome via PR. Please follow these guidelines:

1. Install the pre-commit hooks to ensure your code is formatted correctly.
2. Please keep your PR in "Draft" until it is ready for review.
3. If your PR resolves an issue, please link the issue in the PR description
4. If you can, try running the test suite locally to ensure your changes are working as expected.

Pre-Commit Hooks

Please install the pre-commit hooks to ensure your code is formatted correctly.

uv run pre-commit install

Tests

uv run pytest -v                    # everything
uv run pytest tests/unit -v         # unit only
uv run pytest tests/integration -v  # integration only
uv run pytest -v -m "not gpu"       # CPU-only (inverse of the gpu marker)

License

This project is licensed under the Apache 2.0 license, as found in the License file.

Citation

If you find our work useful, feel free to cite it using

@misc{primeintellect2025prime-rl,
  author = {Prime Intellect},
  title = {prime-rl},
  url = {https://github.com/PrimeIntellect-ai/prime-rl},
  year = {2025}
}