Ernie 4.5

Baidu has open-sourced its ERNIE 4.5 AI models, featuring multi-modal Mixture-of-Experts with enhanced orthogonality. Modong Space, China’s national computing platform, is the first to integrate these advanced models. ERNIE 4.5 optimizes training and inference for improved text generation, image comprehension, and multi-modal reasoning, catering to diverse real-world applications.

Baidu has released its ERNIE 4.5 open-source large language model series, featuring ten models including Mixture-of-Experts (MoE) variants. These models offer fully open-sourced pre-training weights and inference code, accessible via platforms like Hugging Face. The series boasts innovative heterogeneous MoE architecture for multimodal capabilities and achieves state-of-the-art performance on various benchmarks, outperforming competitors in text and multimodal tasks. The models are distributed under the Apache 2.0 license, promoting both academic and commercial use.

Baidu’s Q1 2025 revenue rose 3% year-over-year to ¥32.45B ($4.47B), beating estimates, though non-GAAP net profit fell 8% to ¥6.47B amid increased AI R&D investments. CEO Robin Li emphasized prioritizing applied AI innovations over foundational models, announcing development of Ernie 5.0 and plans to open-source Ernie 4.5 by June 30. The company aims to commercialize digital human technology and scale its autonomous driving unit, Apollo Go, which operates 1,000 driverless vehicles in 15 cities. Ride-hailing orders surged 75% to 1.4M trips in Q1, totaling 11M cumulative journeys. Li projected profitability through cost reductions and operational scaling, positioning Apollo Go as a future growth engine.