New Deepseek Flash model matches OpenAI’s GPT-5.6 Luna at roughly 60 percent lower cost

What happened

Deepseek released a major update to its budget AI model V4 Flash, known as “0731.” This update elevated the model’s score on the Artificial Analysis Intelligence Index by ten points, bringing it to 50. That puts Deepseek Flash just one point behind OpenAI’s GPT-5.6 Luna. The standout detail is that Deepseek achieves this performance gap at roughly 60 percent lower cost per task compared to OpenAI’s model.

Why it matters

The update signals a shift for cost-conscious AI users and businesses. Deepseek Flash’s improved accuracy and efficiency make it a stronger competitor to OpenAI’s premium offerings, especially for workloads sensitive to running expenses. Companies running large-scale AI projects or startups watching tight budgets may find Deepseek’s new model offers a viable high-performance alternative without the premium price tag.

It pressures OpenAI to justify its higher cost given the near parity in performance. Operators deploying generative AI at scale now have a clearer option to reduce AI spend without a significant hit to output quality. This cost-performance trade-off matters in environments where model use is frequent and price per task adds up, such as content generation, summarization, or voice assistants.

What to watch next

It will be important to track how Deepseek scales access and support for this upgraded Flash model, and whether developers adopt it broadly. Competitors like OpenAI may respond with either price adjustments or new feature pushes to defend premium market segments.

Watch how performance holds up in real-world scenarios beyond benchmark scores. Also, seeing if Deepseek expands this pricing advantage to other AI model tiers or develops complementary services could reshape AI economics for smaller players and cost-sensitive operators.

AI Quick Briefs Editorial Desk