LaunchAI ModelsNovember 5, 2026

Baidu releases ERNIE 5.5 Flash Base, a 280B MoE model

Read original source →ernie.baidu.com

The pre-training-only checkpoint activates 14B parameters per token, trained from scratch on 33.5T tokens with a 128K training context and 1M-token context for fine-tuning and inference. Baidu reports MMLU-Pro 78.4 versus DeepSeek-V4-Flash-Base's 69.2, but trails it on HELMET-Rerank, 56.2 to 60.0.

3 sources

More stories today

Open the live feed