Astral-4B

Astral 0.6B is the smallest sized model in the Astral family. It was fine-tuned from Qwen3 0.6b on LucidityAI/Astral-Post-Training-Dataset.

As with usual Qwen3 models, reasoning can be toggled through the usage of /no_think or not.

<|im_start|>user
What is the capital of France?
<|im_end|>
<|im_start|>assistant
<think>

<|im_start|>user
What is the capital of France? /no_think
<|im_end|>
<|im_start|>assistant
<think>

Safetensors

Model size

596M params

Tensor type

BF16

Inference Providers NEW

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

Model tree for LucidityAI/Astral-0.6B-Flash

Base model

Finetuned

Finetuned

(312)

this model

Quantizations