Together AI
Together AI
A cloud platform for building and running generative AI models, providing the fastest inference stack and tools for model training and customization.
Pricing
Tool Info
Rating: N/A (0 reviews)
Date Added: February 20, 2024
Categories
Description
Together AI is a comprehensive cloud platform designed to accelerate the development and deployment of generative AI models. It offers the fastest available inference stack, dedicated GPU clusters for training, and the capability to build custom models. With an emphasis on open-source and privacy, Together AI facilitates model fine-tuning with private data, ensuring users retain full ownership.
Their offerings include Together Inference, Together Fine-tuning, Together Custom Models, and Together GPU Clusters, catering to various needs from model fine-tuning with your data to hosting models for inference and providing high-end compute clusters for large-scale training and fine-tuning tasks
Pricing is highly competitive and varies across services:
Inference pricing is based on the model size and the number of tokens processed, with special promotional pricing for Llama-2 and CodeLlama models.
Dedicated instances for hosting your own model are charged hourly based on the hardware type and model size.
Fine-tuning pricing depends on model size, dataset size, and the number of epochs, with an interactive calculator available on their website to estimate costs. GPU Clusters are offered with state-of-the-art hardware connected over fast networks, tailored for distributed training
Chat, Language, and Code Models
Up to 4B tokens: $0.1 per million tokens
4.1B - 8B tokens: $0.2 per million tokens
8.1B - 21B tokens: $0.3 per million tokens
21.1B - 41B tokens: $0.8 per million tokens
41B - 70B tokens: $0.9 per million tokens
Llama-2 and CodeLlama Models
7B tokens: $0.2 per million tokens
13B tokens: $0.225 per million tokens
34B tokens: $0.776 per million tokens
70B tokens: $0.9 per million tokens
Mixture-of-Experts
8x 7B tokens: $0.6 per million tokens
Embeddings Models
Up to 150M tokens: $0.008 per million tokens
151M - 350M tokens: $0.016 per million tokens
Image Models
Based on image size and steps (512x512 or 1024x1024 pixels):
25 steps: $0.001 to $0.01
50 steps: $0.002 to $0.02
75 steps: $0.0035 to $0.035
100 steps: $0.005 to $0.05
Genomic Models
4.1B - 8B tokens: $2.0 per million tokens
Key Features
- Fastest inference stack
- Privacy and control, ensuring data security and model ownership
- Fine-tuning with private data
- Custom model development
- Dedicated GPU clusters for training
- Open-source model availability
- Cutting-edge optimizations in the Together Training stack like FlashAttention-2 for up to 9x faster training and up to 75% lower costs compared to AWS
Use Cases
- AI model development and deployment
- Custom AI solutions for specific business needs
- High-performance AI research
- Document summarization, code generation, entity extraction, chat, and sentiment analysis capabilities, among others