LLM EXPLORER 57,252 MODELS INDEXED

Astral 4B Preview by LucidityAI

By LucidityAI · 11 downloads
HF Repo ↗ Copy URL

Astral 4B Preview is an open-source language model by LucidityAI. Features: 4b LLM, VRAM: 8.1GB, Context: 256K, License: apache-2.0, LLM Explorer Score: 0.2.

  Chat   Code   Conversational Dataset:nvidia/acereason-1.1-s...   Endpoints compatible   Inference   Qwen3   Reasoning   Region:us   Safetensors   Sharded   Tensorflow
Model Card on HF 🤗: https://huggingface.co/LucidityAI/Astral-4B-Preview 

Astral 4B Preview Parameters and Internals

LLM NameAstral 4B Preview
Repository 🤗https://huggingface.co/LucidityAI/Astral-4B-Preview 
Base Model(s)  Qwen/Qwen3-4b-thinking-2507   Qwen/Qwen3-4b-thinking-2507
Model Size4b
Required VRAM8.1 GB
Updated2026-08-01
MaintainerLucidityAI
Model Typeqwen3
Model Files  5.0 GB: 1-of-2   3.1 GB: 2-of-2
Model ArchitectureQwen3ForCausalLM
Licenseapache-2.0
Context Length262144
Model Max Length262144
Transformers Version4.52.4
Tokenizer ClassQwen2Tokenizer
Padding Token<|endoftext|>
Vocabulary Size151936
Torch Data Typebfloat16
Errorsreplace

Best Alternatives to Astral 4B Preview

Best Alternatives
Context / RAM
Downloads
Likes
Fable Traces256K / 8.1 GB5530209
FastContext 1.0 4B SFT256K / 8.1 GB5735357
GRPO 4 70256K / 8.1 GB60
Qwen3 4B Instruct 2507256K / 8.1 GB3109972911
Lightning 4B256K / 8.1 GB136
FastContext 1.0 4B RL256K / 8.1 GB455961
Qwen3 4B Thinking 2507256K / 8.1 GB291023606
Qwen3 4B Instruct 2507 FP8256K / 5.2 GB110074478
Typhoon2.5 Qwen3 4B256K / 8 GB4898776
Neuron 4B Instruct256K / 8.1 GB3131
Note: green Score (e.g. "73.2") means that the model is better than LucidityAI/Astral-4B-Preview.