Altman Responds to Kimi K3: Partial Speed Tier from OpenAI More Cost-Effective

According to Dynamic Observatory monitoring, OpenAI CEO Sam Altman responded to the dark side of the moon open-source flagship model Kimi K3 on the "Invest Like the Best" podcast. The host referred to it as the most noteworthy AI release of the week and inquired whether it would be a milestone event similar to DeepSeek.

Altman did not directly address whether Kimi K3 is a milestone and did not provide capability comparison results. He emphasized that OpenAI aims to offer the most cost-effective models at different speeds and price points, stating that currently, OpenAI is more cost-effective than Kimi under specific latency requirements. However, he did not specify the exact model, price, or testing conditions.

Altman acknowledged that cheap models and open-weight models will persist in the long term. Some users need to deploy and modify weights themselves, so there will always be a market for open-source models. OpenAI will also distill smaller, cheaper models from its large models.

Regarding concerns about OpenAI models being distilled by competitors, Altman said this is not among his top ten concerns. He believes that OpenAI will have a huge demand for inference in the future, and even if the profit margin is low, the revenue scale will be sufficient to continue to afford expensive model training.