Downloads · 30 days
11
19% of all-time downloads
Astral-1.5-4B-preview-fixed
LucidityAI/Astral-1.5-4B-preview-fixed
Astral-1.5-4B-preview-fixed is a machine learning model from LucidityAI. Use it for the machine learning task on the model card, and read the license before you ship it in a product.
Astral 1.5 4B is the medium sized model in the Astral 1.5 family. It was fine-tuned from Qwen3 4B 2507 Instruct on LucidityAI/Astral-1.5-Post-Train-Dataset.
All-time downloads
59
Public
Parameters
4B
8.1 GB on disk
Likes
0
Public
Repo makeup
Click a slice to open those files.
.safetensors8 GB · 100%
At a glance
- Model type
- qwen3
- Access
- Public
- Created
- Oct 10, 2025
- Updated
- Oct 10, 2025
- SHA
- 0007007e
Base models
Model card
From the Hugging Face model README
Astral-1.5-4B-Preview-fixed
Astral 1.5 4B is the medium sized model in the Astral 1.5 family. It was fine-tuned from Qwen3 4B 2507 Instruct on LucidityAI/Astral-1.5-Post-Train-Dataset.
Note: This is on a different repo than the original preview, as for there was an issue merging the original model, leading it it being highly unreliable.
We reintroduce the reasoning effort selection as seen in Astral 1's preview model, yet not used in Astral 1.
The following modes are available for Astral 1.5:
- low: Generates a smaller reasoning trace
- medium: Generates a decent sized reasoning trace:
- high: Generates a generally unrestricted in size trace
- agent: Genreates no reasoning trace. Intended for agentic use.
Usage example:
<|im_start|>reasoning_level
medium
<|im_end|>
<|im_start|>user
What is the capital of France?
<|im_end|>
<|im_start|>
- Type
- qwen3
- Created
- Oct 10, 2025
- Updated
- Oct 10, 2025