88Z AI, 9 May — Baidu has officially released ERNIE 5.1, significantly reducing pre-training costs by compressing parameter scale. The model performs excellently in various benchmarks, ranking fourth globally and first domestically, while demonstrating strong competitiveness in agent capabilities and math competition scores.
At a glance
- ERNIE 5.1 features significant parameter compression, with pre-training costs at only 6% of similar industry models
- The model ranks fourth globally, with agent capabilities surpassing DeepSeek-V4-Pro
The update
Baidu has officially launched its flagship large language model, ERNIE 5.1. This version features deep architectural optimizations, compressing the total parameter scale to approximately one-third of the previous ERNIE 5.0, with activated parameters at about half. This improvement significantly reduces pre-training costs to just six percent of similar-scale models in the industry. In terms of performance, ERNIE 5.1 ranks fourth globally and first domestically on the Arena search leaderboard with a score of 1223. Furthermore, its agent capabilities surpass DeepSeek-V4-Pro, and it achieved a high score of 99.6 in the AIME26 math competition, second only to Gemini-3.1 Pro, demonstrating leading reasoning and execution levels.
Sources & editorial note
This article is an AI-assisted synthesis by 88Z AI based on the public sources below. Product claims and availability are attributed to their publishers. It does not represent independent testing or first-hand reporting.
- 百度文心Source published
Source dates above refer to the original announcements; the date under the headline is this article’s publication on 88Z AI.