OpenAI Board Chair Comments on Kimi K3: Open-Source Models Not Necessarily Cheaper on July 20

On July 20, OpenAI board chair Bret Taylor commented on cost-efficiency of AI models in a CNBC interview, responding to questions about China's newly released open-source model Kimi K3. Taylor stated that enterprises care less about whether a model is open-source or closed-source, and more about whether the token cost invested delivers sufficient real value.

According to Taylor, open-source models are not necessarily cheaper, as completing the same task may require consuming more tokens. By contrast, frontier models from OpenAI and Anthropic, despite higher per-token pricing, offer superior efficiency and capability. Therefore, evaluating a model's cost-effectiveness requires examining total completion cost and final results, not just per-token price alone.