Baidu launches ERNIE 5.0 preview 1022, beating GPT‑5 on charts and document tasks
When Baidu announced the newest ERNIE model - ERNIE 5.0 Preview 1022 - it seemed to catch a lot of developers off guard. Internally, the team ran a handful of tests that put the model side-by-side with OpenAI’s GPT-5 on things like reading charts and understanding long documents. In those runs, Baidu’s system actually scored a bit higher, which is why early-access users are talking about its surprisingly strong text handling.
The company has started pointing to a few language-specific benchmarks where the numbers look good, but it’s careful not to claim it’s now better than every other AI in every reasoning task. Instead, Baidu frames the results as a sign that the gap between its architecture and the broader field is closing. Their own data hints that the difference is shrinking, especially when the job calls for precise work with structured information.
So, while it’s probably too early to call it a win across the board, the signs point to a narrowing distance - a point that sets up the deeper dive into the findings that follow.
The Preview 1022 variant of ERNIE 5.0, tailored for textual performance, showed even stronger language-specific results in early developer access. While Baidu does not claim broad superiority in general language reasoning, its internal evaluations suggest that ERNIE 5.0 Preview 1022 closes the gap with top-tier English-language models and outperforms them in Chinese-language performance. While Baidu did not release full benchmark details or raw scores publicly, its performance positioning suggests a deliberate attempt to frame ERNIE 5.0 not as a niche multimodal system but as a flagship model competitive with the largest closed models in general-purpose reasoning.
ERNIE 5.0-Preview 1022 looks promising - Baidu says its chart and document understanding scores beat OpenAI’s GPT-5 in internal tests. Still, the company stops short of claiming overall reasoning superiority, noting the model is really tuned for text work. The rollout lands just as OpenAI has pushed out GPT-5.1, which advertises lower token usage and a more chatty tone.
Baidu frames the new model as a step toward global competition, tying it to product upgrades and plans to expand overseas. The tricky part is that the numbers come from Baidu’s own labs; we haven’t seen any independent benchmarks yet. It’s unclear whether these task-specific gains will turn into wider enterprise adoption.
Right now the preview is limited to early developer access, so we can’t say how robust it will be in real-world settings. The announcement certainly shows Baidu’s ambition, but the actual impact will hinge on more testing and how the market reacts. I expect stakeholders will keep an eye on how ERNIE 5.0 fits into Baidu’s existing AI suite.
Meanwhile, developers can start playing with it, but the performance claims still need broader verification.
Further Reading
- ERNIE-5.0-Preview-1022, ready for testing in LMArena! - Baidu Blog
- Baidu advances AI with new Kunlun chips and Ernie 5.0 model - China Daily
- Baidu Prepares to Launch Ernie 5.0 as China's AI Race Intensifies - The AI Insider
- Baidu to launch new reasoning model this month, Ernie 5.0 coming soon - Investing.com
Related Reading
Business & Startups Oracle, NVIDIA deepen tie-up to speed sovereign AI and government digital shift
Business & Startups Skilling programs lag AI; students must prioritize aspiration and depth
Business & Startups EPAM and Cursor partner to scale AI coding for global enterprise customers
Related Topic Baidu opens multimodal AI, claims it beats GPT‑5 and Gemini, runs on one 80GB GPU
LLMs & Generative AI
Related Topic AI tools enable product search, store calls and auto‑buy; Baidu unveils ERNIE 5
LLMs & Generative AI
Common Questions Answered
What specific tasks did Baidu's ERNIE 5.0 Preview 1022 outperform GPT‑5 in according to internal tests?
Baidu's internal evaluations showed that ERNIE 5.0 Preview 1022 achieved higher scores than GPT‑5 on chart interpretation and document comprehension tasks. These two categories were highlighted as areas where the Baidu model clearly outperformed its OpenAI counterpart.
How does ERNIE 5.0 Preview 1022 perform on Chinese‑language benchmarks relative to English‑language models?
The preview variant demonstrates superior results on Chinese‑language benchmarks, surpassing top‑tier English‑language models in those specific tests. Baidu notes that while the model narrows the gap with leading English models, its strongest advantage remains in Chinese textual performance.
Did Baidu release the full benchmark scores for ERNIE 5.0 Preview 1022, and what does the company claim about its overall reasoning abilities?
No, Baidu did not publish the complete benchmark details or raw scores for the preview model. The company also stops short of claiming overall reasoning dominance, emphasizing that the model is primarily tuned for textual performance rather than broad general‑purpose reasoning.
What concurrent development from OpenAI is mentioned in the article, and how does it compare to Baidu’s launch?
The article references OpenAI’s rollout of GPT‑5.1, which promises lower token usage and a friendlier personality. While Baidu highlights its model’s chart and document understanding advantages, OpenAI focuses on efficiency and user interaction improvements, creating a parallel competitive narrative.