Baidu has released its latest open-source foundational model, Wenxin 5.5 Flash Base (ERNIE 5.5 Flash Base). This model features a Mixture-of-Experts (MoE) architecture with 280 billion total parameters, activating 14 billion parameters per token. It was trained on 33.5 trillion tokens and supports context lengths up to 1 million tokens for fine-tuning and inference. The model demonstrates state-of-the-art performance on various benchmarks, including knowledge, reasoning, and coding, surpassing models like DeepSeek-V4-Flash-Base. AI
IMPACT This release offers developers a more efficient and capable foundational model, potentially accelerating the development of specialized AI applications and agents.
RANK_REASON Frontier-lab model release with system card [lever_c_demoted from frontier_release: ic=1 ai=1.0]
- Baidu
- DeepSeek-V4-Flash-Base
- ERNIE 5.5 Flash Base
- Hugging Face
- Hy3-preview-Base
- MiMo-V2.5-Base
- Wenxin 5.5 Flash Base
AI-generated summary · Google Gemini · from 1 sources. How we write summaries →