On November 13, 2025, the 2025 Baidu World Conference kicked off in Beijing, centered around the theme of "Effect Emergence." The conference spotlighted the advancements in domestic chips, upgrades to large - scale models, innovations in AI applications, and global expansion efforts.

At the event, Baidu took the wraps off its latest offerings: the next - generation Kunlunxin M100 and M300 chips, along with the Tianchi 256 and 512 super nodes. The M100 is fine - tuned for large - scale inference scenarios and is set to hit the market in 2026. On the other hand, the M300 is tailored for ultra - large - scale multimodal model training and inference, with its launch scheduled for 2027.

The Tianchi super nodes are slated for release next year. Impressively, a single Tianchi 512 super node has the capacity to train trillion - parameter models.

Baidu also unveiled its native full - modal large model, ERNIE 5.0. This model boasts full - modal understanding and generation capabilities, enabling it to handle multiple information input and output formats, including text, images, audio, and video.

In the realm of AI applications, Baidu introduced the self - evolving super - intelligent agent "Famou" and the Huiboxin "real - time interactive digital human." Additionally, the company is stepping up its efforts to reconstruct AI search.

When it comes to global expansion, Apollo Go has extended its reach to 22 cities across the globe. The Huiboxin digital human is now open to the global market, with its first overseas launch taking place in Brazil.

Analysts are of the opinion that Baidu is emerging as a pivotal player in China's AI chip sector. With its full - stack AI solutions and robust demand for AI, Baidu is well - positioned to become a vital supplier in China's AI industry.