Corrects to remove extraneous word at end of paragraph 2

BEIJING, Feb 27 (Reuters) - China's Baidu 9888.HK will launch the next generation of its artificial intelligence model Ernie in mid-March, which will see improved capabilities in areas such as reasoning, a source with direct knowledge of the matter said.

The upgraded 4.5 Ernie model will also feature enhanced multimodal capabilities, the source said.

Multimodal AI systems are capable of processing and integrating various types of data - like text, video, images, and audio - and can convert content across these formats.

Baidu had said earlier this month it would gradually launch the Ernie 4.5 series in the coming months and officially make it open source from June 30.

Baidu did not immediately respond to a request for comment.

Chinese AI startup DeepSeek's roll-out of AI models which it says is on par with, or even better than, industry-leading models in the United States at a fraction of the cost, have roiled the industry and re-energised the global AI race.

Baidu, one of China's earliest tech giants to launch a ChatGPT-style chatbot, has struggled to gain widespread adoption for its Ernie large language model, despite claiming performance comparable to OpenAI's GPT-4, amid fierce competition.

The emergence of DeepSeek has further challenged the Beijing-based company's AI aspirations.

Baidu CEO Robin Li had previously said that keeping AI models closed-source was the only way forward for development in the sector, but has since pivoted after DeepSeek's success.

Li told analysts during an earnings call this month that Ernie 4.5 will be the company's "best model ever,", acknowledging that DeepSeek's success had prompted the shift towards an open-source approach.

Earlier this week, Alibaba also announced it would make its video- and image-generating AI model, Wan 2.1, open source, further increasing competition in the sector.

(Reporting by Liam Mo and Brenda Goh; Editing by Rachna Uppal)

((liam.mo@thomsonreuters.com))