The new DeepSeek AI model has challenged Anthropic

On August 21, DeepSeek, a Chinese artificial intelligence developer, unveiled an experimental AI model capable of processing visual queries. According to the company, the new model’s capabilities are on par with those of a cutting-edge development by its American competitor, Anthropic.

The new release is an experimental version of the flagship text model DeepSeek-V4-Flash, with the addition of multimodal capabilities. This means the system is capable of analyzing images and screenshots, the company explained.

She claims that in tests of its ability to work independently with images, the new system is nearly on par with Anthropic’s state-of-the-art Opus 4.8 model. Systems of this type are capable of solving complex problems without constant human supervision.

DeepSeek has previously released multimodal models, including the DeepSeek-VL family. However, this new development brings image recognition and task-execution capabilities up to the level of the startup’s flagship product line. Users can interact with it via the DeepSeek API, according to the company’s announcement.

Chinese AI companies are competing fiercely with U.S. developers by offering similar quality at a lower price, Bloomberg notes. However, last week DeepSeek sharply raised the rates for its flagship V4 models—in some cases, by a factor of four. This suggests that the Chinese developer is now prioritizing profitability as it prepares for an IPO, according to Bloomberg. The company may file for a listing in mainland China as early as this year, sources told the agency.

This article was AI-translated and verified by a human editor