China's Baidu Unveils Two New AI Models to Compete with DeepSeek

China's Baidu has launched two new artificial intelligence models as competition in the AI industry intensifies, with one model specifically designed to rival DeepSeek's offerings at half the price.

Baidu announced on Sunday the release of ERNIE X1, a reasoning-focused model that the company claims "delivers performance on par with DeepSeek R1 at only half the price." The company describes X1 as the "first deep thinking model that uses tools autonomously" with "stronger understanding, planning, reflection, and evolution capabilities."

The tech giant also unveiled ERNIE 4.5, which features "excellent multimodal understanding ability" with "comprehensively improved" language capabilities in understanding, generation, logic, and memory. Baidu specifically highlighted the model's "high EQ" and ability to understand internet memes and satirical cartoons.

The announcement comes as Chinese AI startup DeepSeek has disrupted the industry with AI models that reportedly match or exceed U.S. industry-leading models at a fraction of the cost, reinvigorating the global AI race.

According to Baidu's announcement on their X account, pricing for the new models is significantly lower than competitors:

• ERNIE 4.5: Input prices start at $0.55 per million tokens, with output prices at $2.2 per million tokens
• ERNIE X1: Input prices begin at $0.28 per million tokens, with output at $1.1 per million tokens

Baidu stated that its AI chatbot ERNIE Bot "has now been made free to individual users ahead of schedule" with both new models accessible to all users via its official website.

For enterprise users and developers, ERNIE 4.5 is currently available through APIs on Baidu AI Cloud's MaaS platform Qianfan, with ERNIE X1 expected to follow soon.

The company also plans to integrate both models across its ecosystem, including Baidu Search and the Wenxiaoyan app.

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

Reporting by Reuters, Farah Master and the Beijing newsroom; Editing by Lincoln Feast.