Baidu Shifts Ernie AI Model To Open Source

Chinese tech giant Baidu is to shift its AI chatbot to an open source development model, following the success of start-up DeepSeek.

Baidu said the next version of its Ernie large language model (LLM) would be open source from 30 June, in a major shift of strategy.

The company’s founder and chief executive Robin Li has long argued closed-source is the only viable model for AI development.

Baidu also said last week the Ernie Bot service would be free to use from 1 April, ending a 17-month trial at charging a fee for the service.

Free offering

Baidu currently charges 49.90 yuan (£5.49) per month for access to its AI models.

The company was one of the first major Chinese firms to heavily invest in AI following the sensation created by the release of OpenAI’s ChatGPT in November 2022, but Baidu has to date struggled to gain a wide following for its AI tools.

ByteDance’s Doubao chatbot is currently the most popular in China, with 78.6 million active monthly users, while DeepSeek has 33.7 million and Ernie Bot has 13 million, according to January data from Aicpb.com.

“We will gradually launch the Ernie 4.5 series in the coming months and officially open source it from June 30,” said Baidu in a WeChat post.

Li has also changed his rhetoric on open source in recent talks, saying at an event in Dubai last week that the model could accelerate AI adoption.

“If you open things up, a lot of people will be curious enough to try it. This will help spread the technology much faster,” he said, according to a Reuters report.

Baidu’s next-generation Ernie 5 model is reportedly planned for release in the second half of this year.

Open source shift

DeepSeek gained worldwide attention last month after releasing its open source V3 and R1 AI models that it said show performance on par with US rivals, but were developed for a fraction of the cost.

Alibaba’s Qwen models, which are also open source, now power the world’s top 10 open source LLMs, according to AI community Hugging Face.

Meta Platforms also has a significant following with its open source Llama AI models.

OpenAI chief executive Sam Altman told a Reddit question-and-answer session last month that he would personally prefer an open source strategy, but also said such a move was not the company’s “current highest priority”.

The company last week said it would simplify its range of models with the release of GPT-5, creating an interface that would choose the right model for the right task instead of offering the user the option of picking the model themselves.

In a post on social media Altman also said GPT-5 would be offered at a standard intelligence level to free ChatGPT users, while paid users would get a “higher level of intelligence” and access to a wide range of tools including voice interaction, the Canvas interface and AI agents.

The move to simplify its models and make them more accessible was seen by some industry watchers as a response to pressure from DeepSeek.