Disclaimer: This model is for testing purposes only. Raw inference with llama.cpp works but using it with ollama currently doesn't.

This model was trained on a Gemini 2.5 Pro (reasoning) dataset. It is a reasoning model.

It has 21 billion parameters in total and 3 billion activated parameters.

GGUF

Model size

22B params

Architecture

ernie4_5-moe

Hardware compatibility

4-bit

Inference Providers NEW

This model isn't deployed by any Inference Provider. 🙋 Ask for provider support

Model tree for Liontix/ERNIE-4.5-21B-A3B-Thinking-Gemini-2.5-Pro-Distill-GGUF

Base model

Finetuned

Quantized

(1)

this model

Dataset used to train Liontix/ERNIE-4.5-21B-A3B-Thinking-Gemini-2.5-Pro-Distill-GGUF