ERNIE-4.5-300B-A47B
About ERNIE-4.5-300B-A47B
ERNIE-4.5-300B-A47B is a large language model developed by Baidu based on a Mixture-of-Experts (MoE) architecture. The model has a total of 300 billion parameters, but only activates 47 billion parameters per token during inference, thus balancing powerful performance with computational efficiency. As one of the core models in the ERNIE 4.5 series, it is trained on the PaddlePaddle deep learning framework and demonstrates outstanding capabilities in tasks such as text understanding, generation, reasoning, and coding. The model utilizes an innovative multimodal heterogeneous MoE pre-training method, which effectively enhances its overall abilities through joint training on text and visual modalities, showing prominent results in instruction following and world knowledge memorization. Baidu has open-sourced this model along with others in the series to promote the research and application of AI technology
Explore how ERNIE-4.5-300B-A47B's powerful reasoning, multimodal capabilities, and extensive world knowledge can be applied to solve complex, real-world problems efficiently.
Scientific Discovery Acceleration
Accelerate research by analyzing complex datasets, generating and verifying hypotheses, and drafting technical papers with step-by-step reasoning.
Use Case Example:
"Assisted a materials science team by analyzing experimental data and proposing novel alloy compositions, leading to a breakthrough in high-strength, lightweight materials."
Intelligent Code Refactoring
Analyze large codebases to identify logical flaws, suggest architectural improvements, and optimize performance across various programming paradigms.
Use Case Example:
"Pinpointed a critical race condition in a Go microservice, providing a precise, concurrent-safe refactoring solution that significantly improved system stability and throughput."
Multimodal Content Intelligence
Leverage multimodal training to analyze and synthesize information from diverse sources, including text, images, and video, for comprehensive insights and content generation.
Use Case Example:
"Generated a detailed market report by analyzing financial news articles, company earnings call transcripts, and product images, providing a holistic view of market sentiment and product perception."
Strategic Market Insights
Perform multi-step quantitative and qualitative analysis on market data, financial reports, and industry trends to infer causal relationships and generate actionable strategic recommendations.
Use Case Example:
"Developed a comprehensive market entry strategy for a tech startup by analyzing competitor patent filings, social media sentiment, and regulatory documents, identifying untapped niches and potential risks."
Automated Compliance Audits
Audit complex legal documents, regulatory frameworks, and policy guidelines by reasoning through logical dependencies, identifying inconsistencies, and flagging potential compliance issues.
Use Case Example:
"Reviewed a large portfolio of GDPR compliance documents for a multinational corporation, identifying several clauses that conflicted with recent regulatory updates and suggesting precise amendments."
Metadata
Specification
State
Deprecated
Architecture
Mixture of Experts
Calibrated
No
Mixture of Experts
Yes
Total Parameters
300B
Activated Parameters
47B
Reasoning
No
Precision
FP8
Context length
131K
Max Tokens
131K
Compare with Other Models
See how this model stacks up against others.
