Open-source running and training of AI models and LLMs.
This is the 2nd launch from Unsloth. View more
Unsloth Studio
Launching today
Unsloth Studio is an open-source, no-code web UI for training, running, and exporting LLMs locally. It transforms unstructured files into datasets and lets you fine-tune models 2x faster with 70% less VRAM, all without writing complex training scripts.
Free
Launch Team
Flowtica Scribe
Hi everyone!
Unsloth Studio just made fine-tuning models way more intuitive and accessible.
Instead of writing complex training scripts, you now get a clean GUI for the full workflow — dataset management from PDFs/CSVs, training config, real-time monitoring, and even auto dataset recipes. Everything stays local and private.
Unsloth already had a HUGE following for its super-efficient LoRA fine-tuning. Studio turns that into a full platform, lowering the barrier so way more people can experiment, customize, and have fun with their own models.
This is going to accelerate the whole open model scene!
@zaczuo Thanks for posting about this!
Unsloth
@zaczuo Thank you for the support appreciate it!
ngl the best part here is probably that it keeps things local. people are way more interested in experimenting with their own data now, but they still want privacy and don’t always wanna fight with setup for hours just to test something.
also curious, what kind of files are people usually starting with first, pdfs, docs, support chats or something else?
@nayan_surya98 Yes it's fully local! You can add any docs! We also added Python + Bash code execution, Web Search via Duck Duck Go, Claude Code Artifacts so HTML rendering of HTML snippets, auto healing tool calling and more!
Unsloth
@nayan_surya98 You can start with any file, you do not need a dataset to begin with. You can upload PDF, DOCX, CSVS, TXT, code, audio etc into the inference chat and it will render. Same is for the Data Recipes which converts those files into an actual structure dataset
this feels like a pretty important step for open models honestly. a lot of people want to fine tune their own models, but the moment they see scripts, configs and setup headaches, they drop the idea. making the whole flow visual and local could open this up to way more people.
curious, who are you seeing get the most value from studio so far, hobbyists learning this stuff or teams actually training models for work?
Unsloth
@akshay_kumar_hireid Thank you Akshay, we would say it's targeted for a very wide audience, beginners, hobbyists, in production, enterprise, businesses and more :)
Nice local AI plus and no code training hits a growing niche for privacy-conscious devs. I’d lead with the main benefit: Train and run AI models locally, faster and with less VRAM, no coding required.
Trufflow
This could be a really interesting way to help companies better customize their own LLMs based upon their own company's context, processes, and culture. Is use of the Unsloth's Data Recipes the best way to tackle something like this?
great job guys, cant wait to try this
@erildo Thank you! We're open to any suggestions as well :)
Unsloth
@erildo Thank you Erildo appreciate it, hopefully it works great for you. Let us know if you encounter any installation issues :)
Fine-tuning models feels way faster now. Less memory use makes testing multiple setups so much smoother.