Anahtar Kelimeler：ERNIE 4.5, Baidu Açık Kaynak, Çok Modelli Model, PaddlePaddle Çerçevesi, Apache 2.0 Lisansı, ERNIE 4.5 Performansı, MoE Modeli, MFU %47, DeepSeek-V3 Karşılaştırması, Qwen3 Rekabeti
🔥 Spotlight
Baidu releases and open-sources the ERNIE 4.5 model series: Baidu has officially open-sourced its ERNIE 4.5 series, which includes 10 text and multimodal models with parameter sizes ranging from 0.3B to 424B (a MoE model with 47B activated parameters). The series is trained on the PaddlePaddle framework, achieving an MFU of 47%, and is released under the Apache 2.0 license, supporting commercial applications. In several benchmarks, its performance surpasses models like DeepSeek-V3 and Qwen3, and its multimodal version outperforms OpenAI o1 in some evaluations. This release not only provides model weights and code but also opens up API services simultaneously, aiming to promote the widespread application of AI technology in both academia and industry. The community has responded enthusiastically, anticipating its competition with top open-source models like Qwen (Source: huggingface, tokenbender, teortaxesTex, jeremyphoward, QbitAI)