Chinese tech giant Baidu has intensified the ongoing AI battle between China and the United States with the launch of Ernie X1, a new reasoning model that it claims delivers performance on par with DeepSeek R1 at half the price. Simultaneously, Baidu unveiled Ernie 4.5, a multimodal foundation model that, according to the company, outperforms OpenAI’s GPT-4.5 in several benchmarks while costing only 1% of GPT-4.5’s pricing.
To further cement its AI ambitions, Baidu announced it would make its chatbot, Ernie Bot, free to the public starting April 1, ahead of schedule. The company also plans to integrate Ernie 4.5 and X1 into its product ecosystem, including Baidu Search, China’s dominant search engine.
But beyond the product rollout, the launch of Ernie X1 is widely seen as a strategic move in China’s escalating AI race with the US, a rivalry that is shaping the global technology industry.
Register for Tekedia Mini-MBA edition 17 (June 9 – Sept 6, 2025) today for early bird discounts. Do annual for access to Blucera.com.
Tekedia AI in Business Masterclass opens registrations.
Join Tekedia Capital Syndicate and co-invest in great global startups.
Register to become a better CEO or Director with Tekedia CEO & Director Program.
For years, China and the United States have been locked in a fierce battle for AI supremacy, with both nations pouring billions into AI research, semiconductor development, and supercomputing capabilities. The release of Ernie X1 and Ernie 4.5 comes at a critical juncture when US President Donald Trump is pushing aggressive AI strategies to maintain America’s technological edge.
In January, Trump unveiled Stargate, a proposed $500 billion AI and semiconductor initiative aimed at cementing the US’ AI leadership and countering China’s rapid advancements. The program seeks to bolster AI chip manufacturing, expand research in artificial intelligence, and strengthen AI applications across military and commercial sectors.
The past Biden administration implemented sweeping restrictions on AI chip exports to China, limiting the country’s access to Nvidia’s high-performance chips—a key component in training large AI models. However, Chinese firms like Baidu, Alibaba, Huawei, and Tencent have ramped up their domestic chip production efforts and are investing heavily in AI models that can rival OpenAI, Google DeepMind, and Anthropic.
The Cost War in AI Models: OpenAI vs. DeepSeek vs. Baidu
Baidu’s aggressive pricing strategy with Ernie 4.5 and Ernie X1 underpins a key battleground in the AI war: cost efficiency. AI models require significant computational power, and the cost of running these models has been a major bottleneck in global AI adoption.
DeepSeek, a Chinese startup backed by hedge fund High Flyer, disrupted the market in late 2024 with its DeepSeek V3 large language model, followed by DeepSeek R1, a reasoning model introduced in January 2025. According to Bernstein Research, DeepSeek’s models are 20 to 40 times cheaper than equivalent OpenAI models, making them highly attractive to businesses and developers.
Baidu’s Ernie 4.5 follows this trend, with pricing that significantly undercuts OpenAI’s GPT-4.5. Input token prices start as low as RMB 0.004 per thousand tokens, with output tokens priced at RMB 0.016 per thousand tokens. Converted into USD, Baidu’s claims check out, but DeepSeek’s V3 remains the most cost-efficient model in the category.
However, Baidu’s Ernie X1 reasoning model emerges as the cheapest of all, costing just 2% of OpenAI’s o1 model.
Beyond pricing, early user feedback suggests strong performance. Venture partner Alvin Foo of Zero2Launch noted on X, “Been playing around with it for hours, impressive performance.”
China’s AI Strategy: The Open-Source Revolution
Baidu’s AI ambitions align with China’s broader AI strategy, which prioritizes open-source development, self-sufficiency in AI chips, and government-backed funding for AI research.
The success of DeepSeek has demonstrated the power of open-source AI, where making models publicly available significantly boosts adoption. Baidu’s CEO Robin Li acknowledged this shift in a February earnings call.
“One thing we learned from DeepSeek is that open-sourcing the best models can greatly help adoption. When the model is open source, people naturally want to try it out of curiosity, which helps drive broader adoption,” he said.
China has been making rapid AI advancements as it aims to become the world leader in artificial intelligence by 2030. Over the past few weeks, the country has unveiled multiple AI breakthroughs, including Manus, an AI agent, and Alibaba’s open-source model QwQ-32B.
Until now, AI insiders had been eagerly awaiting DeepSeek’s R2 launch, expecting it to set a new benchmark in AI reasoning models. But with Baidu’s Ernie collection now in the mix, the race has taken a dramatic turn.