According to River, enterprises can complete a complex reinforcement learning training run through its API in 15 to 20 minutes without maintaining a dedicated infrastructure team. The company also claims its approach offers two to four times the cost savings of closed-source alternatives.
“The way AI is built today is not how it will be built in the future,” Babuschkin said. “AI should be open, freely available, and affordable. It should feel like it is working for the person using it, not the lab that trained it. We started River to allow people and companies to own their intelligence.”
River's platform is built around giving customers greater control over the models they use. Rather than relying exclusively on general-purpose systems, companies can use the API to train models for their own requirements and then deploy the resulting models directly to production.
The service manages technical components including elastic compute, weight transfers and consistency between sampling and training. River charges according to the tokens consumed during training and inference, rather than requiring customers to pay for unused GPU capacity. Those capabilities are one piece of a broader technology stack River is developing. The company is working on hardware alongside its training infrastructure and personalized AI products, with a longer-term goal of creating AI systems that can continually adapt to individual users.
For now, River is focusing that model-control strategy on developers and enterprises. The company argues that organizations should be able to customize and operate AI around their own needs rather than depend on a single general-purpose model designed for a broad user base. “There is a gap between what AI can do and what most companies actually experience,” General Catalyst Managing Director Marc Bhargava said. “Until now, companies have lacked a cost-efficient way to train, tune, and own custom AI models. River closes this gap, helping any company build models on their own data, tailored to how they actually work.”
Babuschkin brings experience from several major AI labs to the effort. Before founding River, he worked on generative modeling and reinforcement learning at Google DeepMind, contributed to large-scale training at OpenAI and co-founded xAI. River's founding team also includes experience from xAI and Tesla in deep learning and reinforcement learning.
General Catalyst CEO Hemant Taneja framed the investment partly around the development of open-weight AI. “American leadership in AI urgently requires leadership in open weight models, while maintaining a lead in closed frontier models,” Taneja said.
The $1.1 billion financing gives River substantial resources to pursue a strategy spanning model training, infrastructure, products and hardware. Its immediate offering, however, is more focused: making reinforcement learning and fine-tuning of open-weight models accessible through an API, while removing much of the infrastructure companies would otherwise need to operate themselves.
This analysis is based on reporting from businesswire.
Image courtesy of River AI.
This article was generated with AI assistance and reviewed for accuracy and quality.
