LLM fine-tuning/training/evaluation/deployment
Oumi is an open-source Python platform for fine-tuning, evaluating, and deploying open-weight LLMs/VLMs with support for SFT, DPO, GRPO, and multi-engine inference.
Same problem, same approach. Swapping one for another is a config change, not a rewrite.
Choose it when you want a lightning-native pretrain/finetune/serve framework with more model recipes but fewer platform orchestration features.
20+ high-performance LLMs with recipes to pretrain, finetune and deploy at scale.
Choose it for a battle-tested, model-heavy fine-tuning library that is simpler and more mature for SFT/DPO/GRPO if you do not need Oumi's eval/deploy layer.
Unified Efficient Fine-Tuning of 100+ LLMs & VLMs (ACL 2024)
Choose it when out-of-the-box coverage of 600+ LLMs and 300+ MLLMs matters more than Oumi's integrated evaluation and deployment tools.
Use PEFT or Full-parameter to CPT/SFT/DPO/GRPO 600+ LLMs (Qwen3.6, DeepSeek-V4, GLM-5.1, InternLM3, Llama4, ...) and 300+ MLLMs (Qwen3-VL, Qwen3-Omni, InternVL3.5, Ovis2.5, GLM4.5v, Gemma4, Llava, Phi4, ...) (AAAI 2025).
Solves the same problem with a different architecture or at a different layer. Expect to rewrite the integration.
Choose it for low-level distributed PyTorch training when you need direct parallelism control instead of Oumi's high-level multi-engine orchestration.
🚀 Pytorch Distributed native training library for LLMs/VLMs with OOTB Hugging Face support
Choose it for ultra-large MoE training at scale on GPU clusters, as a specialized throughput engine rather than a general end-to-end platform.
A Next-Generation Training Engine Built for Ultra-Large MoE Models
Choose it for TPU/JAX-scale pretraining or fine-tuning when you need native JAX performance instead of Oumi's Python-first multi-framework stack.
A simple, performant, and scalable Jax LLM!
Choose it when you need native Apple MLX fine-tuning on Mac hardware instead of Oumi's PyTorch/cloud-oriented stack.
Fine-tune LLMs on your Mac with Apple Silicon. SFT, DPO, GRPO, Vision, TTS, STT, Embedding, and OCR fine-tuning — natively on MLX. Unsloth-compatible API.
An earlier or more limited way to do the job. Still the right call when you need something small, proven, or CPU-only.
Choose it for YAML-driven fine-tuning that fits an 8B model on a 4 GB laptop GPU, accepting much narrower scope than Oumi.
Fine-tune LLMs from one YAML. Layer streaming trains an 8B model on a 4 GB laptop GPU.
Choose it for a simpler fine-tuning-centric library when you only need basic SFT/DPO without Oumi's deployment and evaluation breadth.
Build, personalize and control your own LLMs. From data pre-processing to fine-tuning, xTuring provides an easy way to personalize open-source LLMs. Join our discord community: https://discord.gg/TgHXuSJEk6
Choose it if you specifically need the old adapter-family PEFT research code for reproducing EMNLP 2023 experiments rather than broad LLMOps.
Code for our EMNLP 2023 Paper: "LLM-Adapters: An Adapter Family for Parameter-Efficient Fine-Tuning of Large Language Models"
Does the same job, but ships as an app. Useful to a person, not swappable into a codebase.
Choose it as an end-user desktop UI for local low-memory fine-tuning/running instead of embedding a library into your codebase.
Local UI to run and train LLMs and diffusion models. Supports GGUF, MLX, Qwen3.8, DeepSeek-V4, MiniMax-H3, Gemma 4, FLUX and more.
Choose it as a visual research environment for training and evaluating models instead of integrating a Python library like Oumi.
The open source research environment for AI researchers to seamlessly train, evaluate, and scale models from local hardware to GPU clusters.
These projects were analysed and named oumi among their alternatives. The relationship is not symmetric — how oumi rates them is a separate judgement, made when oumi is analysed in its own right.
calls oumi “choose when you want a full train/evaluate/deploy orchestration framework instead of fine-tuning only.”
Unified Efficient Fine-Tuning of 100+ LLMs & VLMs (ACL 2024)
calls oumi “Choose when you want a reproducible, YAML-config-driven LLM fine-tune/evaluate/deploy framework independent of the ModelScope ecosystem.”
Use PEFT or Full-parameter to CPT/SFT/DPO/GRPO 600+ LLMs (Qwen3.6, DeepSeek-V4, GLM-5.1, InternLM3, Llama4, ...) and 300+ MLLMs (Qwen3-VL, Qwen3-Omni, InternVL3.5, Ovis2.5, GLM4.5v, Gemma4, Llava, Phi4, ...) (AAAI 2025).
calls oumi “Choose it when you want an all-in-one Python/CLI platform for fine-tuning, evaluation, and deployment rather than a desktop app.”
Local UI to run and train LLMs and diffusion models. Supports GGUF, MLX, Qwen3.8, DeepSeek-V4, MiniMax-H3, Gemma 4, FLUX and more.