AI companies continue to raise huge amounts of capital to fuel their commercial — and, in some cases, open source — ambitions.
I see Togethera startup building open source AI and AI programming infrastructure, which today was announced that it closed a $102.5 million Series A funding round led by Kleiner Perkins with participation from Nvidia and Emergence Capital. The tranche — more than five times the size of the company’s previous round — will go toward expanding Together’s cloud platform that allows developers to build on open and custom models, according to co-founder and CEO Vipul Ved Prakash.
“Startups and enterprises are trying to build an AI strategy for their business without locking into a single vendor,” Prakash wrote in a blog post published on Together’s website this morning. “Open source AI provides a strong foundation for these applications with more and more powerful production models being released almost weekly… We believe genetic AI is a platform technology, a new operating system for applications, and will have a long-term impact on human society. The AI ecosystem will consist of proprietary models and open models, and it’s incredibly important that this future has options and choices.”
Vipul co-founded Together, which launched in June 2022, with Ce Zhang, Chris Re and Percy Liang. Prakash previously founded social media search platform Topsy, which was acquired in 2013 by Apple, where he later became a senior director.
With Together, Prakash, Zhang, Re and Liang seek to create open source models and services that help organizations integrate AI into their applications. To that end, Together has built an open-source cloud platform for running, training and refining models that the co-founders claim offers scalable computing at lower prices than mainstream vendors (e.g. Google Cloud, AWS, Azure).
Currently, Together operates a cloud spanning data centers in the US and EU, including servers operated by partners Crusoe Cloud and Vultr, providing approximately 20 exaflops of computing in total. Customers include Pika Labs, NexusFlow, Voyage AI, and Cartesia, some of which also leverage Together’s model-serving APIs.
“By building custom infrastructure, we can deliver significantly better economics on pretraining and inference workloads,” Vipul continued. “The Together AI platform allows developers to quickly and easily integrate leading open source models or build their own models through pre-training or refinement. Our customers choose to bring their productive AI workloads to Together AI thanks to our industry-leading performance and reliability. while still having the comfort of owning the outcome of their AI investment and always being free to run their model on any platform.”
Together it also invests heavily in AI research. One of the company’s first projects, RedPijama, aims to promote a set of open source artificial intelligence models, including “chat” models according to OpenAI’s ChatGPT. Together also released GPT-JT, a fork of the open source text generation model GPT-J-6B (released by the EleutherAI research group), and OpenChatKit, an attempt at a ChatGPT equivalent.
