Unsloth AI Releases Unsloth Studio: Local No-Code Interface for High-Performance LLM Fine-Tuning with 70% Less VRAM Usage

Unsloth Studio was released, unsloth AI released an open-source local no-code interface for fine-tuning LLMs.

Triton kernels provide 2x faster training speeds, unsloth’s kernels are written in OpenAI’s Triton language and replace generic CUDA kernels.

Unsloth kernels reduce VRAM usage by 70%, specialized kernels reduce memory requirements compared with standard frameworks.

Unsloth Studio enables fine-tuning of 8B and 70B parameter models on one GPU, the Studio supports fine-tuning of models on a single GPU for certain model sizes.

Supports 4-bit and 8-bit quantization via LoRA and QLoRA, unsloth Studio supports parameter-efficient techniques.