Open Source

2x faster training with 70% less VRAM via custom Triton and mathematical kernels — no accuracy loss
LoRA, QLoRA, and full fine-tuning at 4-bit, 16-bit, and FP8 precision
GRPO reinforcement learning support with 80% less VRAM than standard implementations
500K context window fine-tuning for long-document and multi-turn conversation specialization
Supports 500+ models: Llama 4, Qwen 3.5, DeepSeek, Gemma 4, Mistral, Phi-4, gpt-oss
Unsloth Studio web UI for self-hosted inference with GGUF, LoRA, and safetensors formats
Visual dataset recipe editor that ingests PDFs, CSVs, and DOCX files
Multi-GPU training support and free Colab notebooks for getting started

Related Projects

TrendingInference

GitHub

165.0K15.0K

Ollama

ollama

The simplest way to run LLMs locally with 165K+ GitHub stars. One-command deployment, 100+ models, REST API, and multi-platform support.

llama.cpp

ggml-org

Pure C/C++ LLM inference engine supporting CPUs, Apple Silicon, CUDA, and Vulkan

SGLang

sgl-project

SGLang is a high-performance open-source serving framework for LLMs with RadixAttention prefix caching, zero-overhead CPU scheduling, and prefill-decode disaggregation — deployed across 400,000+ GPUs at xAI, LinkedIn, and Cursor.

SGLang

sgl-project

High-performance LLM and multimodal model serving framework with RadixAttention and structured generation.

Apache-2.088

Open Source

Unsloth

Key Features

Tags

Related Projects

Ollama

llama.cpp

SGLang

SGLang