Applied Compute seeks $350M funding as open-source AI competition intensifies

The rise of open-weight AI models is driving new competition in the enterprise sector. Following the impact of DeepSeek’s R1 model, former OpenAI researchers Yash Patil, Rhythm Garg, and Linden Li founded Applied Compute. The startup focuses on training custom open-source models for enterprise clients and hosting them on its own cloud infrastructure.

Applied Compute has already secured clients including Nvidia, Microsoft, and DoorDash. The company is currently in discussions to raise $350 million in funding, which could bring its valuation to $3.25 billion, nearly doubling its worth from four months prior.

In a parallel development in the open-source space, DeepSeek has released an experimental version of its multimodal model, ‘DeepSeek-V4-Flash-Vision-Exp’. Released under an MIT license for free commercial use, the model utilizes a Mixture-of-Experts (MoE) architecture. Benchmarks indicate the model outperforms Anthropic’s Claude Opus 4.8 in certain image agent tasks, such as ‘Agents’ Last Exam’ and ‘ZeroBench’.