The AI platform for creators and builders

From first prompt to production.

Your prompt library, assets, team, and 130+ models. Built for the messy middle between the first idea and thousandth iteration.

Nano Banana 2

LTX-2.3 Pro

GPT-Image 2

Veo 3.1

Kling O1

Seedance 2.0 - Fast

Qwen Image Edit

WAN 2.7

Kling O3 - 4K

Z-Image Turbo

Seedance 2.0 - 4K

Kling O3 - Pro

Trusted by a community of engineers, researchers, and creatives at companies like

Generating is the easy part

Easily switch between the industry's best video generation, image generation, and LLMs. Iterate on prompts and outputs. Ship your own workflows through the API.

Organize your work

As projects scale, keep everything organized. Tag, search, and filter outputs, trace each one back to the prompt, model, and source behind it, and sort into folders with every change tracked.

Team Project

Bravehoof

2 mins ago

Landscapes620 MB

Hero Shots1.34 GB

Status

Prompt

Collaborate with your team

As your team grows, your whole team stays in sync. Every change is saved and versioned automatically, so nothing gets overwritten and you can roll back to any point in time. No more final_v3_REALLY_locked.mp4.

Activity

John labeled 4 images of spotted cows main

2m ago

Maya branched off to landscapes

1hr ago

Priya added 120 background images of hay 🌾

3hr ago

Liam added a hotfix to main

5hr ago

Sofia deleted 32 blurry bovine

8hr ago

Henry merged landscapes into main

1 day ago

🎨 For Creatives

Generate it, organize it, share it.

Bring generative AI into your creative work and manage your growing catalog of assets without the chaos.

Create videos and images, brainstorm with LLMs
Organize and search your growing asset library with tags and filters
Fine-tune a model on your style, no code needed

Start creating

🦾 For Engineers

Prompt it, fine-tune it, ship it.

Building an AI-powered product? Go from zero to scale without managing GPUs, wrangling large datasets, or setting up your own infrastructure.

Use a single API for hundreds of models across every modality
Deploy custom fine-tuned endpoints in minutes
Version millions of files and terabytes of data with full lineage

Read the docs

Train your own models

Train models across any modality: LLMs, vision-language, image, and video to build models that capture your style, your brand, and your domain. Feed your generations and curated datasets straight into fine-tunes.

Fine-tune job

character-lora-flux-v3

Training

Base model

flux klein · 9B

Dataset

bravehoof · 218 rows

Training data

Epoch 4 / 6

67%

Loss

training loss

Scale to Terabytes

Built for terabyte scale. Push, pull, and quickly update terabytes of data in your cloud or ours, without the slowdowns and storage limits that bottleneck other tools. The oxen command line tool is 20x faster than git-lfs.

git

2GB

.txt
.json
.md
.py
.cpp

oxen

20TB+

Images
Video
Audio
Text
Parquet

01 / Versioning

Track every change.

Every dataset, model weight, prompt, and trace keeps its full history of what changed, when, and who changed it. Nothing is lost or overwritten. Any result stays reproducible long after you made it.

See how versioning works

02 / Scale

Built for scale.

Oxen is built for files that are too big for Git. Datasets, model weights, videos, images, 3D models, audio, and more. For when you say "Syncing to S3 will be slow, unless we zip it first. But zipping it will take forever."

View benchmarks

03 / Open Source

Your infrastructure or ours.

Deploy the oxen in your own cloud or ours. We'll help you set it up. Your data and models never have to leave your control to get the full platform. The core is open source and extensible for your own infrastructure.

View on GitHub

MG · Serious Player

Join the growing Herd

Oxen.ai has developed a strong and growing community of researchers, engineers, developers, and creatives. Every Friday we get together and read research and fine-tune models.