Chinese artificial intelligence (AI) startup DeepSeek has accelerated its competition with U.S. AI companies by upgrading its existing model just a month after its initial release.
DeepSeek recently unveiled the 'V4.1-Flash-1017,' an enhanced version of its 'V4.1-Flash,' according to a report by China Business News on October 19. The new model improves processing speed, reduces response length, and enhances overall performance. The upgrades include improvements in inference, coding, and AI agent functionalities. DeepSeek noted that performance has also improved in practical applications such as game development and 3D modeling.
Based on its internal evaluations, DeepSeek claims to have surpassed U.S. competitors in coding capabilities. It achieved a score of 77.0 on the coding performance metric 'DeepSWE v1.1,' outperforming Anthropic's 'Claude Opus 4.5' (74.0) and OpenAI's 'GPT-5.6 Sol' (73.0).
The company also maintains competitive pricing, asserting that its cost per coding task is $0.20, which is one-tenth of the price of U.S. competing models. This upgrade reflects a broader trend in the Chinese AI industry, where competition is expanding beyond model performance to include processing efficiency, cost, and practical industrial applicability.
Additionally, DeepSeek's V4 model is optimized for Huawei's Ascend chips, and it is reported that Huawei chips were used in the training of the V4-Flash. DeepSeek is developing its technology by first securing capabilities with U.S. semiconductors and then enhancing compatibility with Chinese semiconductors.
* This article has been translated by AI.
Copyright ⓒ Aju Press All rights reserved.