Agent frameworks & infrastructure

Models, fine-tuning & training

2,504 agents · most starred first

Models, fine-tuning & training, page 2 of 26

harness-1★ 975

🚀 Ultra Recipe for Training Long-Horizon Search Agents - matching frontier AI's search capability with a 20B model + stateful harness

AgileRL★ 970

The complete post-training stack for LLMs, plus RL for any problem. SFT, DPO and multi-turn agentic RL, with evolutionary hyperparameter optimization built in.

Transformers-for-NLP-2nd-Edition★ 969

Transformer models from BERT to GPT-4, environments from Hugging Face to OpenAI. Fine-tuning, training, and prompt engineering examples. A bonus section with ChatGPT, GPT-3.5-turbo, GPT-4, and DALL-E including jump starting GPT-4, speech-to-text, text-to-speech, text to image generation with DALL-E, Google Cloud AI,HuggingGPT, and more

AgentGym-RL★ 894

Code and implementations for the paper "AgentGym-RL: Training LLM Agents for Long-Horizon Decision Making through Multi-Turn Reinforcement Learning" by Zhiheng Xi et al.

JaxMARL★ 878

Multi-Agent Reinforcement Learning with JAX

Qwen2.5-Coder-7B-Instruct★ 875

transformers safetensors qwen2 text-generation code codeqwen

gpt-tokenizer★ 862

The fastest JavaScript BPE Tokenizer Encoder Decoder for OpenAI's GPT models (gpt-5, gpt-o*, gpt-4o, etc.). Port of OpenAI's tiktoken with additional features.

a-evolve★ 860

The official repository of "Position: Agentic Evolution is the Path to Evolving LLMs".

JaxMARL★ 840

Multi-Agent Reinforcement Learning with JAX

MAAC★ 813

Code for "Actor-Attention-Critic for Multi-Agent Reinforcement Learning" ICML 2019

train-deepseek-r1★ 809

Building DeepSeek R1 from Scratch

Orion★ 808

Orion-14B is a family of models includes a 14B foundation LLM, and a series of models: a chat model, a long context model, a quantized model, a RAG fine-tuned model, and an Agent fine-tuned model. Orion-14B 系列模型包括一个具有140亿参数的多语言基座大模型以及一系列相关的衍生模型,包括对话模型,长文本模型,量化模型,RAG微调模型,Agent微调模型等。

contrastors★ 806

Train Models Contrastively in Pytorch

Qwen-7B-Chat★ 801

transformers safetensors qwen text-generation custom_code zh

optim-agent★ 796

LLM agents as your hyperparameter optimizer.

slimevolleygym★ 793

A simple OpenAI Gym environment for single and multi-agent reinforcement learning

LoongSage★ 785

A production-grade, high-efficiency Agentic RL framework for frontier LLMs with validated recipes.

ToolOrchestra★ 773

ToolOrchestra is an end-to-end RL training framework for orchestrating tools and agentic workflows.

RAG-FiT★ 765

Framework for enhancing LLMs for RAG tasks using fine-tuning.

MADRL★ 756

Repo containing code for multi-agent deep reinforcement learning (MADRL).

Multi-Agent-Reinforcement-Learning-Environment★ 752

Hello, I pushed some python environments for Multi Agent Reinforcement Learning.

Qwen-AgentWorld-35B-A3B★ 750

transformers safetensors qwen3_5_moe image-text-to-text qwen world-model

replit-code-v1-3b★ 748

transformers pytorch mpt text-generation code custom_code

ProphetNet★ 745

A research project for natural language generation, containing the official implementations by MSRA NLC team.

stable-baselines3-contrib★ 745

Contrib package for Stable-Baselines3 - Experimental reinforcement learning (RL) code

DeepSeek-Coder-V2-Instruct★ 719

transformers safetensors deepseek_v2 text-generation conversational custom_code

DeepSeek-Coder-V2-Lite-Instruct★ 718

transformers safetensors deepseek_v2 text-generation conversational custom_code

GPT-NeoXT-Chat-Base-20B★ 694

transformers pytorch gpt_neox text-generation en license:apache-2.0

KAT-Coder-V2.5-Dev★ 691

transformers safetensors qwen3_5_moe image-text-to-text code agent

AutoDidact★ 689

Autonomously train research-agent LLMs on custom data using reinforcement learning and self-verification.

BenchMARL★ 688

BenchMARL is a library for benchmarking Multi-Agent Reinforcement Learning (MARL). BenchMARL allows to quickly compare different MARL algorithms, tasks, and models while being systematically grounded in its two core tenets: reproducibility and standardization.

Llama3-8B-Chinese-Chat★ 688

transformers safetensors llama text-generation llama-factory orpo

DeepCoder-14B-Preview★ 686

transformers safetensors qwen2 text-generation conversational en

Relax★ 678

An Asynchronous Reinforcement Learning Engine for Omni-Modal Post-Training at Scale

stable-code-3b★ 671

transformers safetensors gguf stablelm text-generation causal-lm

Agents-A1★ 671

transformers safetensors qwen3_5_moe image-text-to-text moe vlm

Open-AgentRL★ 664

RLAnything (ICML 2026) & AutoTool (ICML 2026), DemyAgent: Open-Source RL for LLMs and Agentic Scenarios

autonomous-learning-library★ 660

A PyTorch library for building deep reinforcement learning agents.

xLAM★ 643

xLAM: A Family of Large Action Models to Empower AI Agent Systems

Atria-Dawn★ 640
Baichuan-13B-Chat★ 633

transformers pytorch baichuan text-generation custom_code zh

KernelAgent★ 631

Autonomous GPU Kernel Generation & Optimization via Deep Agents

pytorch-DRL★ 619

PyTorch implementations of various Deep Reinforcement Learning (DRL) algorithms for both single agent and multi-agent.

axrl★ 613

AxisRL is an agentic RL post-training framework built on SGLang rollout, Megatron training, and real-world agent workflows.

deepseek-coder-33b-instruct★ 606

transformers pytorch safetensors llama text-generation conversational

zeta★ 605

Build high-performance AI models with modular building blocks

Agent_Foundation_Models★ 589

Chain-of-Agents: End-to-End Agent Foundation Models via Multi-Agent Distillation and Agentic RL.

North-Mini-Code-1.0★ 588

transformers safetensors cohere2_moe text-generation conversational chat

DeepScaleR-1.5B-Preview★ 584

transformers safetensors qwen2 text-generation en dataset:AI-MO/NuminaMath-CoT

Relax★ 580

An Asynchronous Reinforcement Learning Engine for Omni-Modal Post-Training at Scale

practical_seq2seq★ 567

A simple, minimal wrapper for tensorflow's seq2seq module, for experimenting with datasets rapidly

ptan★ 561

PyTorch Agent Net: reinforcement learning toolkit for pytorch

malib★ 558

A parallel framework for population-based multi-agent reinforcement learning.

Llama3-ChatQA-1.5-8B★ 558

transformers safetensors llama text-generation nvidia chatqa-1.5

Atria-Dawn-Preview★ 553
off-policy★ 552

PyTorch implementations of popular off-policy multi-agent reinforcement learning algorithms, including QMix, VDN, MADDPG, and MATD3.

slime-agentic★ 552

A project implementing various agentic RL based on the Slime post-training framework

Deep-RL-Keras★ 548

Keras Implementation of popular Deep RL Algorithms (A3C, DDQN, DDPG, Dueling DDQN)

PokerRL★ 548

Framework for Multi-Agent Deep Reinforcement Learning in Poker

neural-chat-7b-v3-1★ 546

transformers pytorch safetensors mistral text-generation LLMs

MiniMax-M2.1★ 544

MiniMax M2.1, a SOTA model for real-world dev & agents.

aikit★ 544

🏗️ Fine-tune, build, and deploy open-source LLMs easily!

deepseek-coder-6.7b-instruct★ 540

transformers pytorch safetensors llama text-generation conversational

LongCat-Flash-Chat★ 539

LongCat-Flash-Chat safetensors text-generation transformers conversational custom_code

VAGEN★ 524

World model reinforcement learning for multi-turn VLM agents. RL for vision framework (NeurIPS 2025).

gem★ 519

A Gym for Agentic LLMs

agentic-rl-lab★ 519

Reproducing and studying RL algorithms for LLM agents, including GRPO, GSPO, DAPO, OPD, Search-R1, ReTool, ALFWorld and beyond.

GDPO★ 517

Official implementation of GDPO: Group reward-Decoupled Normalization Policy Optimization for Multi-reward RL Optimization

agent-world-model★ 493

Agent World Model: Infinity Synthetic Environments for Agentic Reinforcement Learning

DeepSeek-V2-Chat★ 486

transformers safetensors deepseek_v2 text-generation conversational custom_code

gear★ 483

Toolkits for Recursive Self-improvement (RSI). A modular framework for recursive harness evolution.

Hy3-preview★ 480

Hy3 preview (295B A21B), a leading reasoning and agent model in its size, with great cost efficiency

opal★ 477

OPAL: a reactive optimizer. A coding agent rewrites the optimizer while your model trains.

skill-training★ 475

Skill training with a corruption and reconstruction loop: train an agent's skill library, not its weights.

safety-starter-agents★ 463

Basic constrained RL agents used in experiments for the "Benchmarking Safe Exploration in Deep Reinforcement Learning" paper.

RLAIF-V★ 461

[CVPR'25 highlight] RLAIF-V: Open-Source AI Feedback Leads to Super GPT-4V Trustworthiness

pytorch-rl★ 453

This repository contains model-free deep reinforcement learning algorithms implemented in Pytorch

Harness-RL★ 453

Agentic RAG R1 Framework via Reinforcement Learning

Qwen3.5-4B-GLB-3D-Coder★ 452

safetensors gguf qwen3_5 code threejs 3d

learning-to-communicate★ 447

Learning to Communicate with Deep Multi-Agent Reinforcement Learning

Open-LLaVA-NeXT★ 438

An open-source implementation for training LLaVA-NeXT.

tinyzero★ 435

Easily train AlphaZero-like agents on any environment you want!

Baichuan2-13B-Chat★ 434

transformers pytorch baichuan text-generation custom_code en

salina★ 430

a Lightweight library for sequential learning agents, including reinforcement learning

Stoix★ 425

🏛️A research-friendly codebase for fast experimentation of single-agent reinforcement learning in JAX • End-to-End JAX RL

terminal-bench-rl★ 420

GRPO training code which scales to 32xH100s for long horizon terminal/coding tasks. Base agent is now the top Qwen3 agent on Stanford's TerminalBench leaderboard.

YuLan-Rec★ 419
AgentCPM-Explore★ 418

transformers safetensors qwen3 text-generation conversational en

RL4VLM★ 417

Official Repo for Fine-Tuning Large Vision-Language Models as Decision-Making Agents via Reinforcement Learning

llama2_7b_chat_uncensored★ 416

transformers pytorch tensorboard safetensors llama text-generation

Tree-GRPO★ 412

[ICLR 2026] Tree Search for LLM Agent Reinforcement Learning

tiktoken-rs★ 411

Ready-made tokenizer library for working with GPT and tiktoken

xllm★ 409

🦖 X—LLM: Cutting Edge & Easy LLM Finetuning

genrl★ 405

A PyTorch reinforcement learning library for generalizable and reproducible algorithm implementations with an aim to improve accessibility in RL

SkillZero★ 391

[EMNLP 2026] Official code for "SKILL0: In-Context Agentic Reinforcement Learning for Skill Internalization"

AReno★ 391

An easy-to-use, fast toolkit to scale up RL post-training on a single node.

SkyText-Chinese-GPT3★ 389

SkyText是由奇点智源发布的中文GPT3预训练大模型,可以进行文章续写、对话、中英翻译、内容风格生成、推理、诗词对联等不同任务。| SkyText is a Chinese GPT3 pre-trained large model released by Singularity-AI, which can perform different tasks such as chatting, Q&A, and Chinese-English translation.

decision-forge★ 388

Compile organizational policy into verifiable decision training data.

SDAR★ 383

Official code for "Self-Distilled Agentic Reinforcement Learning"

siiRL★ 382

siiRL: Shanghai Innovation Institute RL Framework for Advanced LLMs and Multi-Agent Systems

More in Agent frameworks & infrastructure