In the AI industry, Together AI recently announced the completion of a $305 million Series B funding round, which has attracted widespread attention. The company's rise is closely related to its newly launched deep reasoning model, DeepSeek-R1. Contrary to initial concerns, many industry experts believe that advancements in deep reasoning have not diminished the demand for infrastructure; rather, they have continuously increased this demand.
Image Source Note: Image generated by AI, image licensed by service provider Midjourney
Since its establishment in 2023, Together AI aims to simplify the use of open-source large language models (LLMs) for enterprises. Over time, the company has gradually expanded its platform, offering a solution called the "Together Platform," which supports the deployment of AI in virtual private clouds and on-premises environments. In 2025, Together AI introduced inference clusters and Agentic AI capabilities, further enhancing the functionality of its platform.
According to Together AI CEO Vipul Prakash, DeepSeek-R1 has a parameter count of up to 671 billion, making its inference running costs significant. To meet the growing user demand, Together AI launched the "Inference Cluster" service, providing clients with dedicated computing power ranging from 128 to 2000 chips to ensure optimal model performance. Additionally, the request processing time for DeepSeek-R1 is typically long, averaging two to three minutes, which has also contributed to the increased demand for infrastructure.
In terms of applications for the inference model, Together AI has identified specific use cases such as coding agents, reducing model hallucinations, and achieving self-improvement of models through reinforcement learning. These applications not only enhance work efficiency but also improve the accuracy of model outputs.
Furthermore, Together AI acquired CodeSandbox to enhance its capabilities in autonomous intelligent workflows. This acquisition enables it to execute code quickly in the cloud, reducing latency and improving the performance of agent workflows.
Faced with fierce market competition, Together AI's infrastructure platform is continuously being optimized, with the deployment of its next-generation Nvidia Blackwell chips, which will provide higher performance and lower latency for model training and inference. Prakash noted that compared to other platforms like Azure, Together AI's inference speed has significantly improved, greatly satisfying customers' demands for high-performance AI infrastructure.
Key Points:
🌟 Together AI secures $305 million in funding to advance deep reasoning models.
📈 The complexity of DeepSeek-R1 significantly increases infrastructure demand, leading to the launch of the "Inference Cluster" service to meet market needs.
🚀 The newly acquired CodeSandbox and Nvidia Blackwell chips will further enhance Together AI's market competitiveness.
