Deepseek V4 Flash 0731 catches GPT-5.6 Luna at 60% lower cost
deepseek v4 flash 0731 matches gpt-5.6 luna on benchmarks but costs 60% less. price-to-performance leader puts pressure on openai.
deepseek has released v4 flash 0731, a major upgrade that scores 50 points on the artificial analysis intelligence index, trailing openai's gpt-5.6 luna by only a single point. despite proximity in quality, the deepseek model operates at roughly 60 percent of the cost, leveraging a 98 percent cache discount and reduced token usage. performance improved across all categories, with agentic tasks on gdpval climbing from 1,189 to 1,559 elo points, while the architecture retains a 1m context window and open mit licensing. this signals a shift where price-to-performance, not just raw capability, determines market leadership.