Unsloth is an open-source AI training platform that enables developers to fine-tune large language models 30x faster and with 90% less memory through custom kernels and optimized training pipelines.
Pricing
Free (open-source)
Reviews
500+
Founded
2024
Team Size
1-10 employees
Current Deal
Free and open-source
About Unsloth
Unsloth provides tools for fine-tuning large language models with dramatically improved speed and efficiency. The platform's custom CUDA kernels and training optimizations enable 30x faster training with 90% less memory usage compared to standard approaches, making it possible to fine-tune models on consumer-grade GPUs.
With over 10 million monthly model downloads and 40,000+ GitHub stars, Unsloth has become the go-to tool for developers who want to create custom AI models. The platform supports popular model architectures including Llama, Mistral, and Gemma.
A Y Combinator S2024 company, Unsloth is building an all-in-one solution for creating highly accurate custom models, with the mission of making model training and open-source AI as accessible as possible.
Pricing
Free (open-source)
Open-source: Free, Apache 2.0 licensed
Pro: Coming soon - Managed training infrastructure
Case Studies
Notable Customers
- AI researchers
- ML engineers
- Startups