What the hell is Deepseek doing for so long?
Chinese AI firm DeepSeek lags behind rivals, still on v3.2 while competitors surge ahead.
The Chinese AI landscape is evolving rapidly, but one notable player, DeepSeek, appears to be stuck in neutral. While competitors like Xiaomi, Baidu (Ernie), Alibaba (Qwen), and others are pushing out increasingly sophisticated models with multimodal capabilities, DeepSeek's flagship model remains at version 3.2, receiving only incremental updates. This stagnation is puzzling given the substantial resources the company reportedly secured following its viral international attention earlier this year. The lack of a significant new release has led industry watchers to question the company's strategic direction and technical roadmap.
A critical gap in DeepSeek's portfolio is the absence of a competitive multimodal model. In an era where leading AI systems from OpenAI (GPT-4V), Google (Gemini), and Chinese counterparts can process and generate images, audio, and video, DeepSeek's text-only focus is seen as a major limitation. This raises fundamental questions about its ability to compete in the broader AI market. Can a company that sparked global interest with a capable, open-source text model survive without evolving into a more versatile, frontier AI lab? The pressure is mounting for DeepSeek to deliver a truly groundbreaking release to reclaim its position.
- DeepSeek's core model remains at version 3.2 with only minor updates, while Chinese rivals release frequent major upgrades.
- The company lacks a publicly released multimodal AI model, a standard feature for frontier AI systems from US and Chinese labs.
- Despite earlier viral success and resource influx, its slow pace casts doubt on its ability to compete long-term.
Why It Matters
The AI race is accelerating; companies that fail to iterate quickly risk immediate obsolescence in a winner-take-most market.