DeepSeek, a Chinese artificial intelligence (AI) company based in Hangzhou, Zhejiang Province, has announced that its latest AI model, DeepSeek V4.1 Flash, is now available for testing. The model is set to officially launch on September 10, 2026, and during the testing phase, it will be priced the same as its predecessor, the V4 Flash version. Price reductions for the V4 Flash are scheduled to take effect on the official launch date. The model is accessible through the DeepSeek API, with a limit of 20 simultaneous requests per account. DeepSeek V4.1 Flash is designed for a wide range of applications, including 3D modeling, video game development, spatial reasoning, and intelligent simulations. It processes up to 427 tokens per second, making it particularly useful for tasks that require speed and efficiency. While it surpasses the V4 Flash version in performance, cost, speed, and task execution time, it still lags behind some of its competitors, such as GPT-6, Astra, and Fable 5.1, in physics-based simulations. Additionally, the model has been noted for a tendency to overthink, which can slow down result generation in certain scenarios. Despite these limitations, DeepSeek V4.1 Flash is gaining attention as a cost-effective and versatile AI tool. It is being tested via API and will be available until its official launch in September 2026. The model is particularly well-suited for professionals who need a balance of speed, affordability, and functionality in their AI workflows. Its ability to handle complex tasks without sacrificing quality makes it a practical choice for industries such as architecture, gaming, and simulation-based research. The pricing structure for DeepSeek V4.1 Flash during the testing phase remains consistent with the V4 Flash version. Outside of peak hours, the unit price is $0.003 for successful cache accesses, $0.15 for missed accesses, and $0.60 for outputs. During peak hours, these rates double. Starting September 10, 2026, DeepSeek plans to adjust the Flash series pricing, and users are advised to plan their usage accordingly. The model’s affordability and performance are expected to make it a valuable resource for a broad range of industries.