Together AI Showcases Surging Growth and New AI Infrastructure at Inaugural AI Native Conf

New updates have been reported about Together AI.

Together AI used its first AI Native Conf in San Francisco to signal rapid commercial and technical momentum, reporting 10x year-over-year growth in annual contract revenue, thousands of enterprise customers, support for over one million developers, 27 contracts above $1 million, and a flagship deal exceeding $1 billion. Positioned as a core infrastructure layer for AI-native companies, Together AI now underpins production-scale inference, pre-training, and model optimization for fast-scaling customers, while its systems research lab directly channels frontier work like FlashAttention and ThunderKittens into deployed cloud services.

At the event, the company introduced FlashAttention 4, which it says can deliver up to 4x performance gains for long-context workloads such as coding agents and complex document reasoning, tightening the gap between theoretical and realized performance. Together AI also launched a Reinforcement Learning API that separates inference from training to enable globally distributed reinforcement learning pipelines, ThunderAgent for open-source, program-aware agent serving with up to 3.6x throughput and lower memory usage, and ATLAS-2, which adapts to real-time user data to provide roughly 1.5x faster inference, collectively expanding its value proposition for customers scaling generative and agentic AI systems.

CEO Vipul Ved Prakash framed the strategy as fusing cutting-edge research with production-grade infrastructure so that the same teams publishing foundational kernels and systems are responsible for the code that runs in customer environments. With generative AI already adopted by an estimated 70% of companies and a new wave of AI-native startups reaching $100 million in annual recurring revenue at unprecedented speed, Together AI aims to be the default cloud for this cohort by offering performance, cost efficiency, and flexibility that can support rapid scaling. The conference, featuring leaders from high-growth AI-native firms and large-scale practitioners, serves as both a demand signal and a platform to deepen Together AI’s ecosystem, positioning the company for continued revenue expansion, larger strategic contracts, and increased influence over the AI infrastructure stack.

Executives evaluating Together AI as a partner or competitor should note the company’s emphasis on open-source tooling, high-performance kernels, and reinforcement learning infrastructure, all oriented toward reducing time-to-production and total cost of ownership for large-scale AI workloads. The combination of billion-dollar-plus contract wins, strong developer adoption, and a consistent pipeline of research-to-production releases suggests Together AI is moving beyond a niche provider into a potential foundational player in AI cloud infrastructure, with implications for hyperscalers, specialized GPU platforms, and AI-native startups that must decide where to anchor their core model operations.