DeepSeek's overwhelming cost-effectiveness...expected to create an even bigger impact than R1

DeepSeek launched V4 Pro and V4 Flash, positioning both as low-cost open-source models with performance close to top-tier systems like Opus 4.7 and GPT-5.5.

The models use a large MoE architecture, long context, and massive training data, helping them deliver strong results on benchmarks at much lower cost.

Analysts say the pricing could make DeepSeek especially attractive to enterprise buyers focused on cost efficiency and self-hosted deployment.

The release is also intensifying debate over geopolitical risk, Chinese open-source dependence, and whether U.S. export controls are still effective.