Baidu Unveils Ernie X1, an Open-Source AI Reasoning Model
Baidu Launches New AI Models: Ernie X1 and Ernie 4.5
Baidu, a prominent Chinese technology company often likened to Google, has introduced two innovative AI models named Ernie X1 and Ernie 4.5. These models are poised to compete with similar offerings from major players in the AI space, notably OpenAI and DeepSeek.
Introducing Ernie X1 and Ernie 4.5
On a recent Saturday, Baidu unveiled Ernie X1, a reasoning model designed to provide superior performance at a significantly lower cost. Baidu claims that Ernie X1 rivals the performance of DeepSeek’s R1 model while being offered at just half the price. Alongside this, the company also launched Ernie 4.5, a multimodal foundation model that reportedly outperforms OpenAI’s GPT-4.5 across multiple evaluation metrics, while costing merely 1% of that price.
Pricing and Cost-Effectiveness
Baidu’s strategic pricing for these models underscores the competitive landscape of AI development. For instance, the cost for input and output tokens for Ernie 4.5 starts at around 0.004 Chinese yuan per thousand input tokens and 0.016 per thousand output tokens. When these rates are converted to U.S. dollars, Ernie 4.5 costs approximately $0.55 per million input tokens and $2.20 per million output tokens. In contrast, OpenAI’s GPT-4.5 charges about $75 and $150, respectively, for the same amounts, showing a dramatic difference in pricing.
- Ernie 4.5 Token Pricing:
- Input Tokens: $0.55 per million
- Output Tokens: $2.20 per million
Reasoning Models: Another Competitive Edge
When it comes to reasoning models, Baidu’s Ernie X1 also shows considerable cost benefits. It is priced at roughly $0.28 per million input tokens and $1.10 per million output tokens, making it less than 2% of the cost of OpenAI’s equivalent models.
- Ernie X1 Token Pricing:
- Input Tokens: $0.28 per million
- Output Tokens: $1.10 per million
This shift in pricing comes amidst heightened scrutiny of AI model costs, especially as companies like DeepSeek continue to enter the market with competitive prices.
Industry Impact and Open-Source Initiatives
Baidu’s latest offerings signify a more extensive trend within the AI industry, leaning towards open-source development. The rise of companies like DeepSeek—launching models that are significantly cheaper than their OpenAI counterparts—has highlighted the benefits and effectiveness of open-source practices in AI.
Baidu’s CEO, Robin Li, pointed out the advantages of making advanced models accessible to the public. He noted that open-sourcing high-performing models tends to foster wider adoption, as curiosity drives potential users to experiment with them.
Growing Competition in AI
The emergence of Ernie X1 and Ernie 4.5 underscores the competitive dynamic between AI firms in both the United States and China. Baidu is not the only player seeking to gain an upper hand; they follow closely behind advancements like DeepSeek’s V3 and R1 models, which have recently garnered significant attention for their functionalities.
Furthermore, the Chinese government aims to position the country as a global leader in AI by 2030. Recent releases, including AI agents like Manus and Alibaba’s QwQ-32B model, reflect this ambition.
User Feedback
Feedback from AI enthusiasts who have explored Ernie’s capabilities has been largely favorable. Alvin Foo, a venture partner at Zero2Launch, expressed his satisfaction by saying, "Been playing around with it for hours, impressive performance." This highlights that the practical performance of Baidu’s models is meeting the expectations set by their pricing.
As the landscape of AI technology evolves, Baidu’s latest models may shift the competitive balance, giving other firms a reason to reassess their strategies and offerings in this fast-paced market.
