Together AI
By Together AI
Together AI is an AI infrastructure company that provides an “AI acceleration cloud” for training, fine‑tuning, and serving open and custom models on large GPU clusters, targeting AI‑native companies that want better performance and economics than general‑purpose clouds. It positions itself as a leading platform for open‑source and enterprise AI, serving hundreds of thousands of developers and hosting hundreds of models across text, vision, and multimodal use cases.
Together AI offers a full stack for generative AI: GPU cloud infrastructure, model APIs, fine‑tuning and pre‑training pipelines, and enterprise deployment options (public cloud, VPC, and on‑prem). The platform is marketed as delivering ~30% or better performance and cost efficiency compared to traditional cloud solutions for AI workloads, with customers running trillion‑token inference and large‑scale training jobs on its clusters.
The company was founded in 2022 by CEO Vipul Ved Prakash along with co‑founders Ce Zhang, Chris Ré, and Percy Liang, who bring backgrounds from Topsy, Apple, and leading ML research labs. Since launching, Together AI has grown from a small research‑driven team into a nearly 200‑person organization headquartered in San Francisco, expanding revenue from an estimated 15 million USD in 2023 to about 50 million USD in 2024, with projections of 120 million USD in 2025. Its platform now serves over 450,000 AI developers and provides access to 200+ open‑source models, reflecting rapid ecosystem adoption.
Together AI has raised approximately 534 million USD in total funding across multiple rounds. A 20 million USD seed round in 2023 led by Lux Capital funded its initial open‑source model and cloud platform push. The company closed a 102.5 million USD Series A in 2023 led by Kleiner Perkins with participation from Nvidia and Emergence, aimed at scaling its generative AI cloud. In March 2024 it raised 106 million USD (often referenced as Series B1) led by Salesforce Ventures, reaching unicorn status; by February 2025 it secured a further 305 million USD Series B co‑led by General Catalyst, Coatue, and Salesforce Ventures, bringing total funding above 530 million USD and valuing the company at about 3.3 billion USD.
Together AI’s core product is an AI cloud that exposes a large library of open and specialized models (e.g., Llama, DeepSeek, Qwen, Mixtral and others) via APIs, often compatible with OpenAI’s interface, so customers can swap providers with minimal code changes. It provides performance‑optimized inference with systems like ATLAS and Turbo for speculative decoding, as well as full and lightweight fine‑tuning, retrieval‑augmented generation patterns, and large‑scale pre‑training support on its GPU clusters. Enterprise offerings include private/VPC deployments, compliance-oriented controls, and the Together Enterprise Platform, which lets organizations run generative AI workloads in their own cloud or on‑prem with Together’s orchestration, observability, and optimization stack.
Together AI describes itself as a research-driven, open‑source–oriented AI company that wants to make powerful generative models broadly accessible and to establish open alternatives to closed AI ecosystems. Culture summaries emphasize transparency, innovation, and societal impact, with a strong focus on open‑source collaboration, curiosity‑driven learning, and resource optimization; the environment encourages experimentation and cross‑functional teamwork. Employee-facing materials highlight flexible, often remote-friendly roles, comprehensive benefits, and a mission-driven mindset centered on advancing open AI infrastructure rather than purely proprietary stacks.
Together AI serves a broad base of AI‑native startups and larger enterprises seeking high‑performance, open‑model infrastructure; reports note that it already surpasses 100 million USD in annualized revenue with plans to double its workforce to meet demand. The platform is used by more than 450,000 AI developers and supports over 200 open‑source models, becoming a key hub for developers who want to run Llama, DeepSeek, and other models without managing GPU fleets themselves. Community engagement comes through open‑source releases (e.g., GPT‑JT, OpenChatKit, Red Pajama), research publications, developer content, and a startup accelerator program that offers credits and engineering support to AI-native companies, further reinforcing Together AI’s role as both infrastructure provider and ecosystem catalyst.
