Dohyeon1/ERNIE-M-SMoE-ngroups56

TEXT GENERATIONPricing:Input $0.32 / Cached $0.016 / Output $1.6Concurrent Unit Cost:1Model Size:21BQuant:FP8Context Size:32kPublished:Sep 19, 2026Architecture:Transformer Featherless Exclusive Cold

Dohyeon1/ERNIE-M-SMoE-ngroups56 is a 21 billion parameter Sparse Mixture-of-Experts (SMoE) model, likely based on the ERNIE-M architecture, designed for efficient processing with a context length of 32768 tokens. This model leverages a grouped SMoE design, suggesting optimizations for performance and scalability in large language model applications. Its architecture is intended for tasks requiring substantial contextual understanding and efficient inference.

Loading preview...

Model Overview

This model, Dohyeon1/ERNIE-M-SMoE-ngroups56, is a 21 billion parameter Sparse Mixture-of-Experts (SMoE) model. While specific details regarding its training data, architecture, and intended use cases are marked as "More Information Needed" in the provided model card, its designation as an ERNIE-M-based SMoE model with 56 groups implies a focus on efficient and scalable language processing. The model supports a substantial context length of 32768 tokens, indicating its capability to handle long sequences of text.

Key Characteristics

  • Architecture: Sparse Mixture-of-Experts (SMoE) with 56 groups, likely building upon the ERNIE-M framework.
  • Parameter Count: 21 billion parameters, positioning it as a large-scale language model.
  • Context Length: Supports a significant context window of 32768 tokens, suitable for tasks requiring extensive contextual understanding.

Potential Use Cases

Given its SMoE architecture and large context window, this model is potentially well-suited for:

  • Applications requiring efficient inference with large models.
  • Tasks benefiting from processing long documents or conversations.
  • General natural language understanding and generation tasks where scalability is a concern.