Together AI

About Together AI

Together AI key features

• **High-Performance Inference**: Offers serverless and batch inference options, optimized for various workloads with dedicated model and container inference capabilities.
• **Accelerated Compute**: Provides reliable GPU clusters and custom infrastructure for scalable AI applications, including instant clusters and optimized performance.
• **Managed Storage**: Features high-performance storage solutions for securely storing model weights and data, with zero egress fees.
• **Custom Training and Fine-Tuning**: Supports model shaping through reinforcement learning and fine-tuning, allowing users to tailor models to specific data and requirements.
• **Developer Environments**: Includes sandbox environments for building and testing AI applications, along with comprehensive documentation and practical implementation guides.

Together AI use cases

• **High-Performance Inference**: Provides serverless and batch inference options for efficient processing of AI models.
• **Custom Model Deployment**: Supports dedicated model and container inference for tailored hardware and model requirements.
• **Managed Storage Solutions**: Offers secure storage for model weights and data, optimized for AI workloads.
• **Model Fine-Tuning**: Enables fine-tuning of open-source models to improve accuracy and control behavior without managing infrastructure.
• **Developer Environments**: Facilitates the creation of sandbox environments for building and testing AI applications.

Useful for

• High-performance inference capabilities through serverless and batch processing options.
• Custom training and fine-tuning of models for specific workloads and improved accuracy.
• Scalable GPU clusters for accelerated compute and efficient resource management.
• Managed storage solutions for secure and optimized data handling.
• Developer environments and tools for building and deploying AI applications effectively.

Price

• **Serverless Inference**: High-performance inference as APIs with no infrastructure management or long-term commitments.
• **Batch Inference**: Cost-effective processing of massive workloads asynchronously, scaling to 30 billion tokens per model.
• **Provisioned Throughput**: Token-based pricing with committed inference capacity and a 99% uptime SLA.
• **Dedicated Model Inference**: Deploy models on dedicated infrastructure for speed, control, and optimal economics.
• **Managed Storage**: High-performance storage for AI workloads with zero egress fees.