The DeepSeek-V4-Pro-0813 model, quietly launched this Wednesday (12) by Chinese startup DeepSeek, took 12th place on the Vals Index. The result places the model behind rivals such as OpenAI's GPT-5.5 and Anthropic's Claude Opus 5.
On the Artificial Analysis Intelligence Index, the model scored 53 points, the same result as Zhipu AI's GLM-5.2. The performance was four points below OpenAI's GPT-5.6 and seven points below Moonshot AI's Kimi K3.
Reception and limitations
The version was released quietly. Some developers were disappointed with the general capabilities and the price, while researchers said they were impressed with the performance in niche areas, such as cybersecurity.
