AMD Developer · Jul 2026
AMD and Unsloth announced official AMD GPU support for training, fine-tuning, reinforcement learning, and deployment across Windows, WSL, and Linux, including ROCm-backed workflows.
Open-Source Reinforcement Learning (RL) & Fine-tuning for LLMs.
Company research
Public company, workplace, funding, and market signals
Updated Jul 29, 2026
Overview
Unsloth AI is a 2023 San Francisco-based open-source AI company focused on making LLM fine-tuning, reinforcement learning, and local inference faster and more memory-efficient. Its public-facing product is the Unsloth library and Unsloth Studio, a local/offline UI for training, running, and deploying models with no-code workflows, tool-calling, and broad model/hardware support.
Primary product
Unsloth Core + Unsloth Studio for local LLM fine-tuning, reinforcement learning, and inference
Founded
2023
Headquarters
San Francisco, California, United States
Team size
11-50
Work style
Flexible
Industry
Technology, Information and Internet
Sub-industry
Open-source AI software for LLM fine-tuning, reinforcement learning, and local model serving
Offices
San Francisco, California 94114 (HQ)
Business model
open-source
freemium
pro tier
enterprise licensing
Funding
Stage
seed
Total raised
$540K
Latest round
Seed · Sep 2024
Sep 2024 · Pioneer Fund
Sep 2024 · Y Combinator
May 2024
May 2024 · GitHub Accelerator
People & investors
Work & culture
Open-source, community-driven, and engineering-heavy. Public communications emphasize rapid iteration, model efficiency, bug fixing, collaboration with the open-source ecosystem, and making AI more accessible.
Work style
Flexible
Compensation
Public hiring signal for a Founding ML Engineer: $200K-$500K base salary plus 0.30%-0.70% equity; the role notes new grads are okay and can start remote before becoming in-person in San Francisco.
Product & market
Pricing
Freemium/open-core with paid Pro, Max, and Enterprise tiers; enterprise pricing appears quote-based.
Differentiators
- —Open-source core with paid Studio/Pro/Enterprise tiers
- —Local/offline model training and inference
- —Custom Triton kernels for speed and lower memory use
- —OpenAI-compatible and Anthropic-compatible APIs
- —Tool-calling, web search, and agent workflows in Studio
- —Broad model coverage across Llama, Mistral, Qwen, DeepSeek, Gemma, Phi, and others
- —Cross-platform support for macOS, Windows, Linux, and WSL
- —Support for NVIDIA and AMD hardware
Technology
Python
TypeScript
PyTorch
Hugging Face Transformers
Hugging Face TRL
Triton
llama.cpp
Docker
CUDA
ROCm
Cloudflare
GGUF
Safetensors
Customers
Canva
LinkedIn
NASA
NVIDIA
Competitors
Fireworks
Goodfire
Cartesia
Aleph Alpha
Covariant
AltaML
Scale & contact
Latest news
Unsloth blog · May 2026
Unsloth said it joined the PyTorch Ecosystem Landscape while remaining an independent open-source project; the post describes the product, its kernels, and broader model support.
Unsloth AI Substack · Feb 2026
The update introduced 12x faster MoE training, embedding model support, and longer-context reinforcement learning improvements, alongside a forthcoming UI push.
Unsloth blog · Sep 2024
Unsloth announced YC backing, a studio/waitlist update, and a milestone of roughly 2 million monthly downloads at the time.
VentureBeat · May 2024
VentureBeat covered GitHub Accelerator’s cohort and highlighted Unsloth as a participant focused on faster, lower-memory model fine-tuning.
Jobs and careers at Unsloth AI
0 jobs at Unsloth AI
No open positions
Check back later for new openings