Together AI, a company focused on open-source AI, has made significant strides in securing funding and outlining its mission:
Series A Funding: Together AI raised $102.5M in a Series A financing round1.
Lead Investors: The round was led by Kleiner Perkins, with participation from NVIDIA and Emergence Capital1.
Mission: The company is committed to creating a comprehensive cloud platform that enables developers to build on open and custom AI models1.
Growth Plans: The funding will accelerate the goal of creating the fastest cloud platform for generative AI applications1.
Adoption and Integration: Since its launch, Together AI’s products have seen significant adoption, and the company plans to scale services and introduce new products for AI integration.
Together AI has been advancing its technology development with a focus on creating a fast and efficient cloud platform for generative AI. Here are some key highlights:
Inference and Fine-Tuning: They offer serverless APIs for inference and fine-tuning, which have attracted over 45,000 registered developers1.
GPU Clusters: Dedicated training clusters are available with the Together Training stack, ready to go for users1.
Custom Models: The platform enables the building of state-of-the-art models from scratch with ease1.
Research Innovations: Their research team has contributed to breakthrough AI models, datasets, and optimizations. Notable projects include:
Cocktail SGD: This optimization reduces networking overhead by up to 117x, addressing a major challenge in training generative AI models in a distributed environment1.
FlashAttention 2: Used by all Transformer models, it accelerates training and fine-tuning of LLMs by up to 9x and achieves 72% model FLOPs utilization for training on NVIDIA A100s1.
Sub-quadratic Model Architectures: In collaboration with Hazy Research, they’re developing the next core architecture for generative AI models that will provide faster performance with longer context1.
RedPajama Project: Enables leading generative AI models to be available as fully open-source, with the RedPajama dataset being used to create over 500 leading models1.
Together AI’s platform supports a variety of open-source models and is integrated into AI application development frameworks like LangChain, Vercel, LlamaIndex, MongoDB, and EmbedChain2. This positions them as a key player in the open-source AI infrastructure, providing tools and services that enable rapid integration and development of AI applications.