Markets
StatsAPI
  • Market
  • Wallet
  • News
  1. News
  2. /
  3. large-language-models
  4. /
  5. Baidu Unveils ERNIE 5.0, Claiming Superiority Over GPT-5
post-main
AIlarge language modelsMultimodal Models

Baidu Unveils ERNIE 5.0, Claiming Superiority Over GPT-5

DA
Daniel Reed
2 hours ago7 min read
In a move that signals the intensifying global AI arms race, Chinese tech giant Baidu has unveiled ERNIE 5. 0, its next-generation foundation model, claiming it achieves parity or even superiority against Western counterparts like OpenAI's GPT-5 and Google's Gemini 2.5 Pro. The announcement, made at Baidu World 2025, comes mere hours after OpenAI's own incremental update to GPT-5.1, setting the stage for a direct confrontation in the enterprise AI market. ERNIE 5.0 is architected as a natively omni-modal model, a significant technical distinction that means it is designed from the ground up to jointly process and generate content across text, images, audio, and video within a single, unified architecture. This stands in contrast to the common industry approach of stitching together separate, modality-specific models in a post-hoc fusion, a method Baidu frames as inferior for true contextual understanding.The company is strategically releasing this powerhouse as a proprietary model, accessible only through its ERNIE Bot platform and Qianfan cloud API for enterprises, while simultaneously pursuing an open-source strategy with models like the recently released ERNIE-4. 5-VL-28B-A3B-Thinking under a permissive Apache 2.0 license. This two-track approach allows Baidu to court both large corporate customers needing premium, high-capability services and the developer community hungry for customizable, unrestricted models.The benchmark results presented by Baidu are audacious, suggesting ERNIE 5. 0 outperforms or matches its rivals in critical areas like multimodal reasoning, document understanding, and image-based question answering.It reportedly achieves leading scores on specialized enterprise-focused benchmarks such as OCRBench, DocVQA, and ChartQA, which test a model's ability to parse, comprehend, and reason about information in documents and charts—a core capability for automating financial analysis and legal document review. On image generation, Baidu's internal evaluations claim it ties or exceeds Google's Veo3 in semantic alignment and quality, while its audio understanding capabilities appear robust, if less emphasized.The model's pricing on the Qianfan platform positions it squarely in the premium tier, with input costs at $0. 85 per million tokens and output at $3.40, making it more affordable than Anthropic's Claude Opus but competitive with GPT-5. 1.This pricing underscores a deliberate segmentation in Baidu's model portfolio, differentiating high-volume, low-cost options like ERNIE 4. 5 Turbo from this high-capacity flagship designed for complex, multimodal tasks.CEO Robin Li's statement that the goal is to internalize AI as a 'native capability' transforming 'intelligence from a cost into a source of productivity' reflects a broader industry shift from mere tooling to deeply integrated intelligence. However, the launch was not without immediate real-world hiccups; a prominent AI evaluator publicly flagged a persistent bug where the model would uncontrollably invoke tools during specific tasks, a issue Baidu's developer relations team acknowledged and pledged to fix within hours, demonstrating a new level of responsiveness as it courts a global audience.Beyond the model itself, Baidu's international push is in full swing, with the global rollout of its no-code builder MeDo, the productivity workspace Oreate, and the expansion of its digital human platform, which saw massive adoption during China's recent shopping festival. The scale of Baidu's ambition is further evidenced by its autonomous ride-hailing service, Apollo Go, which now claims the title of the world's largest robotaxi network.This release is more than just a product update; it is a strategic declaration that Chinese AI firms are no longer playing catch-up but are now launching credible, top-tier alternatives to the established Western incumbents. The race is no longer just about raw parameter counts or simple benchmark leads, but about architectural elegance, multimodal fluency, and building a comprehensive ecosystem that can scale globally.While independent, third-party verification of Baidu's performance claims is still pending, ERNIE 5. 0 undeniably marks a pivotal moment, proving that the future of AGI will be shaped by a fiercely competitive, multipolar landscape.
#lead focus news
#Baidu
#ERNIE 5.0
#GPT-5
#multimodal AI
#enterprise AI
#benchmark performance
#global expansion

Stay Informed. Act Smarter.

Get weekly highlights, major headlines, and expert insights — then put your knowledge to work in our live prediction markets.

Comments

Loading comments...

© 2025 Outpoll Service LTD. All rights reserved.
Terms of ServicePrivacy PolicyCookie PolicyHelp Center
Follow us:
NEWS