ERNIE 5.0
Baidu's 2.4T-parameter natively omni-modal flagship for text, image, audio, and video
About
ERNIE 5.0 is Baidu's flagship model, unveiled at Baidu World 2025. It is a natively omni-modal system with up to 2.4 trillion parameters that jointly models text, images, audio, and video in one architecture rather than bolting modalities onto a text backbone. Baidu positions it against GPT-5-class and Gemini-class models, citing 40-plus benchmark tests where its language and multimodal understanding sit near the top, and image/video generation it calls world-leading in professional use. A January 2026 snapshot topped the LMSYS Arena text leaderboard at 1,451 points, per third-party coverage. It is proprietary: access runs through the ERNIE app for consumers and Baidu's Qianfan cloud platform for developers, with reported pricing around $0.85 per million input tokens. Baidu leans on its knowledge graph and Chinese-language data as differentiators. The announcement was met with some market skepticism about differentiation, and independent verification of the headline benchmark claims remains thin.
Specs
Context window
Pricing
Free quota
API access
Weights
Released
Benchmarks
Vendor-stated: competitive with Gemini 2.5 Pro and GPT-5-High across 40+ tests; 1,451 LMSYS Arena text (third-party reported, Jan 2026)
Provider-reported unless a source says otherwise. Benchmarks are marketing-adjacent by nature — treat them as directional.
Use cases
Trust score — 7/10
Scored 7/10 (metrics tier, metrics-v1). Evidence: Baidu World 2025-11-13 launch verified via PR Newswire/Nasdaq; 2.4T params and omni-modal claims are vendor-stated; Arena score third-party reported; pricing secondary-reported.