Baidu

ERNIE 5.1

Baidu

ERNIE 5.1 is Baidu’s more efficient agent and reasoning model derived from ERNIE 5.0’s elastic sub-network training. It compresses total and active parameter capacity while applying asynchronous reinforcement learning and multi-teacher distillation. The hosted model targets language reasoning, creative writing, search, and tool-driven workflows. The official May 9 launch post confirms chat and AI Studio access. Absolute parameter counts and a supported context ceiling are omitted because the announcement establishes relative reductions rather than exact specifications.

ERNIE 5.0

Baidu

ERNIE 5.0 is Baidu’s unified multimodal foundation model, integrating text, images, video, and audio in an autoregressive architecture. It has 2.4T total parameters and was trained from scratch with sparse expert computation and elastic sub-model training. Public preview checkpoints appeared in November 2025, followed by an upgraded release in January and a February technical report. This page groups routine preview snapshots and later service revisions under the major family identity. Public weights are not offered.

ERNIE 4.5

Baidu

ERNIE 4.5 is Baidu’s Apache-licensed family of language and vision-language models. This page groups its dense and sparse text checkpoints with separately labeled multimodal variants. The largest vision-language checkpoint uses 424B total parameters and 47B active parameters, with 128K context. Heterogeneous expert routing supports cross-modal learning, while smaller models offer lower serving costs. The June 30 weight release follows earlier hosted availability; checkpoint-specific modality and size differences must be preserved.