How AI Startups Are Evaluating The Latest Model Advancements

DeepSeek’s R1 is shaking up the AI landscape. Launched on January 20, this advanced reasoning model claims performance on par with OpenAI’s o1 — at just 2% of the cost. Unlike other frontier models reliant on high-end chips and massive datasets, R1 is optimized for older hardware and leverages novel reinforcement learning techniques, allegedly slashing training costs to $6 million.

Beyond efficiency, R1’s open-source MIT license…