131K
0.0936
0.3718
Chat
Active
ERNIE 4.5
The model family includes the reasoning-focused 21B Thinking variant, the standard 21B model, and the high-capacity 300B model, each optimized for different workloads and deployment scales.
.svg)
ERNIE 4.5 excels in multilingual reasoning, cost-efficient inference, and flexible integration for enterprise and research applications.
Baidu ERNIE 4.5 is a new generation of large language models designed for high-quality text generation, structured reasoning, and long-context understanding. Built on a Mixture-of-Experts architecture, ERNIE 4.5 delivers strong performance across a wide range of workloads while remaining computationally efficient and developer-friendly.
The ERNIE 4.5 family includes multiple model sizes and reasoning-focused variants, allowing teams to choose the optimal balance between cost, speed, and intelligence.
ERNIE 4.5 is available in several variants, each targeting a specific class of use cases—from lightweight reasoning to large-scale enterprise workloads.
Optimized for Deep Reasoning & Complex Tasks
Input: $0.0936 per 1M tokens
Output: $0.3718 per 1M tokens
Balanced LLM for General Purpose Text
Input: $0.0936 per 1M tokens
Output: $0.3718 per 1M tokens
High-Capacity Model for Throughput & Scalability
Input: $0.39 per 1M tokens
Output: $1.482 per 1M tokens
High-Capacity Model for Enterprise-Grade Workloads
Input: $0.09295 per 1M tokens
Output: $0.3718 per 1M tokens
Lightweight Preview Model for Rapid Development
Input: $0.858 per 1M tokens
Output: $3.289 per 1M tokens
The Thinking variant prioritizes deeper reasoning and logical accuracy. It is tuned to perform better on tasks that require planning, analysis, and stepwise problem solving.
Standard ERNIE 4.5 models focus on fast, reliable text generation and conversational performance. They are ideal for everyday AI use cases where speed and fluency matter more than deep reasoning depth.
Across the ERNIE 4.5 family, Baidu reports strong results on core text benchmarks:
Baidu ERNIE 4.5 is a new generation of large language models designed for high-quality text generation, structured reasoning, and long-context understanding. Built on a Mixture-of-Experts architecture, ERNIE 4.5 delivers strong performance across a wide range of workloads while remaining computationally efficient and developer-friendly.
The ERNIE 4.5 family includes multiple model sizes and reasoning-focused variants, allowing teams to choose the optimal balance between cost, speed, and intelligence.
ERNIE 4.5 is available in several variants, each targeting a specific class of use cases—from lightweight reasoning to large-scale enterprise workloads.
Optimized for Deep Reasoning & Complex Tasks
Input: $0.0936 per 1M tokens
Output: $0.3718 per 1M tokens
Balanced LLM for General Purpose Text
Input: $0.0936 per 1M tokens
Output: $0.3718 per 1M tokens
High-Capacity Model for Throughput & Scalability
Input: $0.39 per 1M tokens
Output: $1.482 per 1M tokens
High-Capacity Model for Enterprise-Grade Workloads
Input: $0.09295 per 1M tokens
Output: $0.3718 per 1M tokens
Lightweight Preview Model for Rapid Development
Input: $0.858 per 1M tokens
Output: $3.289 per 1M tokens
The Thinking variant prioritizes deeper reasoning and logical accuracy. It is tuned to perform better on tasks that require planning, analysis, and stepwise problem solving.
Standard ERNIE 4.5 models focus on fast, reliable text generation and conversational performance. They are ideal for everyday AI use cases where speed and fluency matter more than deep reasoning depth.
Across the ERNIE 4.5 family, Baidu reports strong results on core text benchmarks:
-min-p-130x130q80.png)
.png){kind=small}
.png){kind=small}
