ERNIE 5.0
Baidu·
Videoflagshipcloud
Overview
2.4T-parameter ultra-sparse MoE flagship in public preview via ERNIE Bot. Natively omnimodal (text/image/audio/video jointly modeled from pretraining). #1 Chinese model on LMArena Text and Vision (#8 globally). China-market only; no Western public API.
Capabilities and innovations
TextImageAudioVideoOmnimodal reasoningAgentic workflowsNatively Omnimodal Unified Autoregressive FrameworkUltra-Sparse MoE (<3% activation per inference)PaddlePaddle Framework
Benchmarks
| Benchmark | Score | Source |
|---|---|---|
| MMLU | 85% | unsourced |
- GPQA Diamond: not reported by the vendor
- Humanity’s Last Exam (no tools): not reported by the vendor
- SWE-bench Verified: not reported by the vendor
- SWE-bench Pro: not reported by the vendor
- MMLU-Pro: not reported by the vendor
- Terminal-Bench 2.x: benchmark did not exist at release
- MMMU-Pro: benchmark did not exist at release
Architecture and hardware
- Parameters
- 2400B total, 72B active per token (MoE)
Links
More from Baidu
Data curated by AI Model Timeline. See the methodology for admission criteria, benchmark eras and source priority.