LucidityAI/Astral-4B-Preview

TEXT GENERATIONPricing:Input $0.4 / Cached $0.08 / Output $0.8Concurrent Unit Cost:1Model Size:4BQuant:BF16Context Size:32kTool Calling:SupportedPublished:Aug 28, 2025License:apache-2.0Architecture:Transformer0.0K Open Weights Featherless Exclusive Cold

LucidityAI/Astral-4B-Preview is a 4 billion parameter language model built upon the Qwen3-4b-thinking-2507 foundation, specifically fine-tuned on the nvidia/AceReason-1.1-SFT dataset. This model specializes in high-fidelity, step-by-step reasoning with configurable depth, making it ideal for tasks requiring logical coherence and structured problem-solving. It is designed for research, evaluation, and development, focusing on enhancing reasoning capabilities.

Loading preview...

Astral-4B-Preview: A Reasoning-Focused LLM

LucidityAI's Astral-4B-Preview is a 4 billion parameter language model engineered for advanced reasoning tasks. Built on the Qwen3-4b-thinking-2507 foundation, it has been extensively fine-tuned using the nvidia/AceReason-1.1-SFT dataset to significantly enhance its logical coherence, problem-solving abilities, and structured thinking.

Key Capabilities & Features

  • Specialized Reasoning: Designed to generate high-fidelity, step-by-step reasoning traces.
  • Configurable Reasoning Depth: Users can specify the desired level of reasoning (none, low, medium, high, ultra) in the system prompt using the Reasoning-level: directive, allowing for tailored output verbosity.
  • Enhanced Problem-Solving: Optimized for tasks requiring detailed logical progression and structured thought processes.

Usage & Status

This model is currently in a preview release phase, intended for research, evaluation, and development. Users must include a valid Reasoning-level: in the system prompt for defined behavior. Feedback is actively encouraged to guide future iterations. It is not yet recommended for production applications.

Finetunes
1 models