Deepseek has released its latest Flash model, V4 '0731,' which marks a significant upgrade to its budget AI offering. According to the Artificial Analysis Intelligence Index, the new version scores 50 points, ten points higher than the previous V4 Flash launched in April 2026. This places it just one point behind OpenAI's GPT-5.6 Luna, despite costing about 60 percent less per task, even after OpenAI's 80 percent price cut. A major factor in this cost efficiency is Deepseek's 98 percent cache discount, which exceeds the industry-standard 90 percent. The model also uses 12 percent fewer tokens than its predecessor. The Artificial Analysis Intelligence Index shows Deepseek V4 Flash '0731' scoring 50 points after its update, nearly matching OpenAI's GPT-5.6 Luna while claiming the top spot for price-to-performance ratio. | Image: Artificial Analysis
The model demonstrates improvements across all tested categories, with the most notable gains in agentic tasks. On GDPval, a benchmark assessing models on complex real-world office work, it scores 1,559 Elo points, up from 1,189. It also hallucinates less frequently. The architecture remains unchanged, featuring 284 billion total parameters, 13 billion active, and a one-million-token context window. The model weights are available under an MIT license on Hugging Face.
Source: thedecoder
