Products
Serverless Inference
API for inference on open-source models
Dedicated Endpoints
Deploy models on custom hardware
Fine-Tuning
Train & improve high-quality, fast models
Evaluations
Measure model quality
Together Chat
Chat app for open-source AI
Code Execution
Code Sandbox
Build AI development environments
Code Interpreter
Execute LLM-generated code
Tools
Which LLM to Use
Find the ‘right’ model for your use case
Models
See all models →
Clusters of Any Size
Instant Clusters
Ready to use, self-service GPUs
Reserved Clusters
Dedicated capacity, with expert support
Frontier AI Factory
1K → 10K → 100K+ NVIDIA GPUs
Cloud Services
Data Center Locations
Global GPU power in 25+ cities
Slurm
Cluster management system
GPUs
Solutions
Customer Stories
Testimonials from AI pioneers
Startup Accelerator
Build and scale your startup
Enterprise
Secure, reliable AI infrastructure
Why Open Source
How to own your AI
Industries & Use-Cases
Scale your business with Together AI
How Hedra Scales Viral AI Video Generation with 60% Cost Savings
When Standard Inference Frameworks Failed, Together AI Enabled 5x Performance Breakthrough
Developers
Documentation
Technical docs for using Together AI
Research
Advancing the open-source AI frontier
Model Library
All our open-source models
Cookbooks
Practical implementation guides
Example Apps
Our open-source demo apps
Videos
DeepSeek-R1: How It Works, Simplified!
Together Code Sandbox: How To Build AI Coding Agents
Pricing
Pricing Overview
Our platform & GPU pricing.
Inference
Per-token & per-minute pricing.
LoRA and full fine-tuning pricing.
GPU Clusters
Hourly rates & custom pricing.
Questions? We’re here to help!
Talk to us →
Company
About us
Get to know us
Values
Our approach to open-source AI
Team
Meet our leadership
Careers
Join our mission
Resources
Blog
Our latest news & blog posts
Events
Explore our events calendar
Knowledge Base
Find answers to your questions
Featured content
Learn how Cursor partnered with Together AI to deliver real-time, low-latency inference at scale
Best practices to accelerate inference for large-scale production workloads
