Baidu’s benchmark table compares the newly released ERNIE-5.5-Flash-Base against DeepSeek-V4-Flash-Base at 284B total and 13B active parameters, close to ERNIE’s 280B total and 14B active. Baidu published 30 benchmarks across seven categories under a single pipeline with identical prompts and few-shot examples against three base models: DeepSeek-V4-Flash-Base (284B-A13B), Hy3-preview-Base (295B-A21B) and MiMo-V2.5-Base (310B-A15B). ERNIE-5.5-Flash-Base has 280B total parameters, 14B active per token, 33.5T pretraining tokens, 128K training context, FP8 and BF16 weights available now, and the ERNIE 5.5 Flash Base Model License Agreement v1.0; DeepSeek-V4-Flash-Base has 284B total, 13B active, 128K-class training context and an MIT licence. Baidu highlights ERNIE results including MMLU 93.3, MMLU-Pro 78.4, GPQA 60.2, BBH 89.2 with chain of thought, HumanEval+ 82.3, MBPP+ 81.5, HELMET-RAG 71.5 and Rep