Ernie 4.5 is Baidu’s frontier AI model with native multimodal abilities
It can understand text, images, audio, and video content
Ernie X1 is Baidu’s first reasoning-focused AI model
Baidu claims that Ernie X1 model will be available at half the price of DeepSeek-R1
Photo Credit: Reuters
Baidu released two new artificial intelligence (AI) models on Sunday. First, the Ernie 4.5 is a foundation model that succeeds the company's previous generation, while the second, Ernie X1, is a reasoning-focused model. The latter is also the Chinese tech giant's first attempt at a “deep-thinking” AI model, and Baidu claims that the large language model (LLM) performs on-par with the DeepSeek-R1. Alongside the model releases, Baidu also made its AI platform Ernie Bot free to access, and users will not have to pay a premium to use the chatbot.
Baidu Releases New Ernie AI Models
In a press release, the Chinese tech giant announced the release of the two new AI models. Currently, only the Ernie 4.5 model has been released and is accessible via the chatbot platform, and the company said that the Ernie X1 will be available soon. Typically, AI companies introduce new models with a research paper and repository (if they're open-source), however, Baidu did not release either.
The company highlighted that the Ernie 4.5 will be open-sourced on June 30 with source code and model weights made available to developers. No such announcement has been made about the reasoning model.
Ernie 4.5 is a natively multimodal AI model with the ability to understand text, images, audio, and videos. This means users can upload these file formats as input and ask the AI queries about them. The company also said that the language capabilities of the model are being enhanced and it can now better respond to conversational messages as well as reasoning and memory-based queries. Ernie 4.5 can also understand contextual content such as memes, satire, and others.
Ernie 4.5 benchmark scores Photo Credit: Baidu
As for performance, Baidu claimed that its latest foundation model outperforms GPT-4.5 across various benchmarks, and is priced at one percent of GPT-4.5. The performance metrics are based on the company's internal testing. Apart from the AI platform, Ernie 4.5 is also available via application programming interfaces on Baidu AI Cloud's MaaS platform Quianfan.
On pricing, Baidu said that Ernie 4.5 input prices start at CNY 0.004 (roughly Rs. 0.04) per thousand tokens, and the output prices are set at CNY 0.016 (roughly Rs. 0.19) per thousand tokens.
Coming to Ernie X1, the company claimed that the multimodal reasoning model comes with the capability of tool use and excels in parameters such as Chinese knowledge Q&A, literary creation, manuscript writing, dialogue, logical reasoning, and complex mathematics. Some of the tools it supports include advanced search, image understanding, document understanding, image generation, code interpretation, webpage reading, TreeMind mapping, and more.
The company said that its first reasoning model was trained using progressive reinforcement learning (RL), an end-to-end training approach, and a unified multi-faceted reward system. Just like DeepSeek-R1, it also comes with a visible chain of thought (CoT).
While Ernie X1 is not available on the Quianfan Platform, once it is released, its input prices will be kept at CNY 0.002 (roughly Rs. 0.02) and output prices at CNY 0.008 (roughly Rs. 0.09) per thousand tokens, respectively.
Baidu’s MuseStreamer AI Video Generation Model Takes on Google’s Veo 3 With Native Audio Support: Report4 July 2025
Search
![[Sponsored] Students Try the Galaxy Tab S10FE for the First Time - Real Reactions, Powerful Features](https://c.ndtvimg.com/2025-07/54d9ikeg_galaxy_160x120_18_July_25.jpg?downsize=180:*)
