# InternLM/xtuner

A Next-Generation Training Engine Built for Ultra-Large MoE Models

Repository: https://github.com/InternLM/xtuner
Canonical: https://ross.abutalabs.com/products/xtuner
Homepage: https://xtuner.readthedocs.io/zh-cn/latest/
Language: Python
License: Apache-2.0
License Family: permissive
Topics: llm, agent, deepseek-v3, gpt-oss, internvl, kimi-k2, multimodal, qwen3-moe, reinforcement-learning, intern-s1, qwen3-vl
Last push: 2026-08-26T11:19:16+00:00

## Health v2 (maintenance only)
Score: 67/100 (v2, computed 2026-09-02T17:46:02.011165+00:00)
- activity 99, release rhythm 16, longevity 82
- inputs: {"age_days": 1149, "days_push": 7, "days_rel": 418, "gap_med": 139, "n_releases_24m": 2}
- flags: none
- formula: round(0.45*activity + 0.35*rhythm + 0.20*longevity); archived -> min(score, 10)

## Adoption (not part of the score)
Stars 5183, forks 445 (observed 2026-08-28T04:09:11.825032+00:00)

## What it is
XTuner is an open-source LLM training engine from InternLM designed for fine-tuning ultra-large-scale Mixture-of-Experts (MoE) models, with dropless training and long-sequence support. It also serves as a flexible toolkit for parameter-efficient fine-tuning of large language and multimodal models.

## Use cases
- fine-tune a large language model on custom data
- train 200B-scale MoE models efficiently
- fine-tune multimodal models like InternVL or Qwen-VL
- train LLMs with 64k sequence lengths
- run parameter-efficient fine-tuning like LoRA
- reinforcement learning training for LLMs

## When to choose
- you need to fine-tune large or MoE-based LLMs like DeepSeek-V3, Qwen3-MoE, or Kimi-K2
- you want memory-efficient long-sequence training without complex 3D parallelism
- you need a flexible open-source toolkit supporting both SFT and RL training
- you work with multimodal models such as InternVL or Qwen3-VL

## When to avoid
- you only need inference or serving rather than training
- you need traditional dense-model distributed training with full 3D parallelism at extreme scale
- you want a no-code GUI training tool

## Facets
- artifact type: library
- maturity: active
- function: llm-training, machine-learning, deep-learning, rag
- domain: large-language-models, deep-learning, machine-learning, artificial-intelligence
- platform: python
- tags: moe, fine-tuning, parameter-efficient-fine-tuning, multimodal, reinforcement-learning, deepseek-v3, qwen3, internvl, long-context, gpu, linux, docker

## Member repositories
- InternLM/xtuner (main) score 67

## Provenance
- Observed fields: from GitHub, fetched 2026-08-28T04:09:11.825032+00:00.
- Health v2: computed from the inputs above; adoption is never an input.
- Inferred fields (summary, facets, guidance): AI-extracted, prompt v1, taxonomy v1, on 2026-08-29T18:01:47.237083+00:00, confidence not recorded.
  - readme: https://github.com/InternLM/xtuner (fetched 2026-08-28T04:09:11.825032+00:00, sha 7539f638da98)
  - registry_pypi: https://pypi.org/pypi/xtuner/json (fetched 2026-08-29T08:56:06.859552+00:00, sha b50a1a741322)
- Data as of 2026-08-30T08:39:29.467469+00:00.