DeepSeek has officially released its latest AI model, V4.1-Flash-1017, on October 17, 2026, marking a significant advancement in artificial intelligence technology. The Chinese company claims this new version is not only faster but also leaner than its predecessors, setting a new benchmark in the AI landscape.
This release comes in the wake of a five-hour outage earlier in the week, sparking discussions about the company’s infrastructure and resource availability. Nonetheless, the launch of V4.1-Flash-1017 has generated excitement within the tech community, as it is now live on DeepSeek’s website, mobile app, and developer platform.
According to the company, the new model boasts a score of 77.0 on the DeepSWE v1.1 benchmark, a notable increase from the 74.2 achieved by its earlier iteration. This improvement reflects DeepSeek’s commitment to enhancing performance and efficiency across various applications.
Key Features of V4.1-Flash-1017
DeepSeek has described this update as a “Pareto improvement,” meaning it significantly enhances both performance and cost-effectiveness. The model is designed to excel in reasoning, coding, and agentic tasks, as well as visual benchmarks, making it a versatile tool for developers and businesses alike.
Utilizing reinforcement learning with self-feedback, V4.1-Flash-1017 is optimized for integration with tools such as DeepSeek Harness, Tencent CodeBuddy, and WorkBuddy. This allows users to create projects from inception to completion with greater efficiency.
In a strategic move to broaden accessibility, DeepSeek has opened its Search and Fetch APIs to all users, priced at $0.75 per 1,000 calls. This initiative is expected to democratize access to high-quality AI tools, making them available to a wider audience.
Benchmark Performance
Aside from the DeepSWE v1.1 score, V4.1-Flash-1017 also achieved a score of 60.5 on the Automation-Bench V1.0.6, up from 54.8 for its predecessor. Furthermore, in the rigorous Humanity’s Last Exam, which tests knowledge across various domains, the new model scored between 36.8 and 42.6, compared to the previous range of 36.8 to 39.1.
These improvements solidify DeepSeek’s position as a formidable competitor against established U.S. players like Anthropic and OpenAI, as well as other Chinese companies such as Moonshot Kimi and Z.ai. Notably, DeepSeek’s open-weight models come at a significantly lower cost than their proprietary U.S. counterparts, which continue to dominate the market.
The earlier version, V4.1-Flash, launched in early September, saw a rapid decline in the stock prices of competitors, including MiniMax Group, Z.ai, and Alibaba Group, highlighting the disruptive potential of DeepSeek’s innovations.
As the AI landscape evolves, DeepSeek’s latest offering is poised to challenge existing paradigms, emphasizing performance improvements while maintaining affordability. The company’s commitment to advancing AI technology remains evident as it pushes boundaries and sets new standards in the industry.
