ERNIE 4.5 21B A3B

Baidu

A sophisticated text-based Mixture-of-Experts (MoE) model featuring 21B total parameters with 3B activated per token, delivering exceptional multimodal understanding and generation through heterogeneous MoE structures and modality-isolated routing. Supporting an extensive 131K token context length, the model achieves efficient inference via multi-expert parallel collaboration and quantization, while advanced post-training techniques including SFT, DPO, and UPO ensure optimized performance across diverse applications with specialized routing and balancing losses for superior task handling.

Try Now

Capabilities

Tool Use

Technical Specifications

Context Window

120,000 tokens

Max Output

8,000 tokens

Pricing

Token Costs (per 1M tokens)

Cache Miss Input

$0.07

Non-Reasoning Output

$0.28

Tool Costs (per 1K calls)

Web Search

$15

Code Execution

$0.19

Legacy

Made legacy on

Reason

21B MoE; limited availability outside China

Recommended Replacement

Qwen3.6 Plus