Baidu, one of China’s tech giants, has just put two new AI models on the market – ERNIE 4.5 and X1. These new tools come at a time when the AI field in China is full of strong rivals all trying to win the top spot.
What’s New With Baidu’s AI Tools
On Sunday, Baidu made its big move by showing off two AI models that stand out for what they can do. The first one, ERNIE 4.5, works with many types of content at once. It can handle text, pictures, sound, and video, and even change content from one type to another.
ERNIE 4.5 isn’t just smart – Baidu says it also has “high EQ.” This means it can get jokes, memes, and even grasp the point of funny cartoons that most AI tools would miss. Its skills with words have grown, and it’s now much better at tasks that need deep thought and good recall.
But the real star might be X1, the other new AI model. Baidu made this one to think in a more step-by-step way. The firm says X1 is the first AI that can pick and use tools on its own, with no help from a person. It can plan what it needs to do, think about how well it’s doing, and get better as it works.
The Price War Starts
What makes these new AI tools stand out isn’t just what they can do – it’s how much they cost.
Baidu claims that ERNIE X1 works just as well as a model from DeepSeek (a key rival), but costs only half as much to use.
For those who build apps and sites, the price points are worth a look. ERNIE 4.5 starts at just $0.55 per million tokens when you feed data in, and $2.20 per million tokens for what it gives back. X1 costs even less at $0.28 per million tokens for input and $1.10 per million tokens for output.
These low prices could make a big dent in how much it costs for firms to use top AI tools in their work and apps.
Why This Matters in China’s AI Race
The news comes right when the AI field in China is at a key point. Not long ago, a firm called DeepSeek shook things up when it put out AI models that it says work as well as the best ones from the US, but cost much less.
This has sparked what some now call an AI price war. As costs drop and what these tools can do gets better, more firms and app makers will be able to add smart AI to what they make and sell.
Baidu isn’t new to AI – they’ve worked on it for close to 20 years. They have a big team that works just on AI tasks and have put lots of cash into this field for a long time. The tech giant was also one of the first in China to make a chat tool like ChatGPT.
Yet, Baidu has had a hard time with its past AI tools. Its ERNIE large language model didn’t catch on as much as the firm had hoped, though Baidu claims it works just as well as OpenAI’s GPT-4. Now, with these new tools, Baidu hopes to change that.
What Sets These Tools Apart
ERNIE 4.5 stands out for how it can work with text, pics, sound, and more all at once. When an AI can work with and mix all these types of data, it can do more kinds of tasks. It can look at a pic and tell you what it shows, turn text into pics, or make clips based on what you type.
But there’s more to it than just that. Baidu says ERNIE 4.5 is now much less prone to make things up (a big flaw in many AI tools). It’s also much better at tasks that need clear, step-by-step thought.
X1 takes things a step past that. It aims to think like a smart person would, by first making a plan, then doing the task, then thinking about how well it did. This helps it learn and get better each time it does a task. It can also pick tools to help it do a job, just like a pro would pick the right tool for each part of a big job.
How You Can Use These AI Tools
For most of us, Baidu has made its ERNIE Bot free to use. You can try both new AI models through the bot on its site at yiyan.baidu.com.
If you run a firm or make apps, you can add ERNIE 4.5 to your own tools right now through Baidu AI Cloud’s MaaS site, called Qianfan. X1 will be there soon, too.
Baidu has more plans in store as well. They want to add these AI tools to lots of their own apps and sites, like Baidu Search, the Wenxiaoyan app, and more.
The Road Ahead: Snags and Next Steps
Not all the news is good, though. Some folks have found bugs with how you sign in, and it’s hard for those not in China to use the tools well. There’s no app for phones that use Android yet, and some find the marks it puts on pics (known as a “watermark”) to be a pain.
Some ask if the tool will be “open source,” which would let folks see how it works and change it to meet their needs. Baidu has not yet said if this will be the case.
For now, all eyes are on how well X1 works in the real world. Some pros ask if it will just be a cheap stand-in for tools like DeepSeek’s R1, or if it can rise to the top of the field in its own right.
Why This News Counts for Tech Pros
If you work in tech, this news could change how you use AI in your work. With the cost of top AI tools now half of what it was, you may be able to add AI smarts to more of what you do.
The choice of which AI tool to use may now hinge more on what each one does best, not just which one you can best fit in your funds. As each new tool tries to stand out in some way, you’ll have more to pick from to match your needs.
For those who run firms, the race to sell the most AI tools means you can now add AI help to more of what you do, with less risk to your bank. The cost to test AI tools in your work now sits at a point where more firms can try it out.
If you teach or learn AI, these new tools add more to look at and work with. You can now try to grasp how each one works, and what makes some think in ways more like how we do.
And if you just like to see what’s new in tech, the race to make the best AI tools now runs at full steam. You can watch as firms push the bounds of what AI can do, and how it might one day be a part of more of what we do each day.
The work to make AI tools that can think like we do, help us in our work and home life, and cost less than past tools now takes a big step. Baidu’s new moves show they want to still be a big name in this field, and that they plan to fight hard to stand out as the best.
What do you think of these new AI tools? Have you had a chance to try them yet? Share what you’ve found in the comments, or tell us how you might use them in your own work.
