CrystalReasoner/Qwen2.5-3B-CrysReas-CrystalTextLLM

Hugging Face
TEXT GENERATIONConcurrency Cost:1Model Size:3.1BQuant:BF16Ctx Length:32kPublished:May 21, 2026Architecture:Transformer Warm

CrystalReasoner/Qwen2.5-3B-CrysReas-CrystalTextLLM is a 3.1 billion parameter language model developed by CrystalReasoner, built on the Qwen2.5 architecture. This model is specifically fine-tuned for generating crystal structures from natural language instructions. It incorporates supervised fine-tuning, thinking traces for crystallographic and physical priors, and reinforcement learning with verifiable rewards. Its primary strength lies in producing valid, stable, and property-conditioned crystal structures.

Loading preview...

CrystalReasoner/Qwen2.5-3B-CrysReas-CrystalTextLLM Overview

This model, developed by CrystalReasoner, is an end-to-end large language model (LLM) framework designed for the generation of crystal structures based on natural language descriptions. Built upon the Qwen2.5 architecture, it features 3.1 billion parameters and a context length of 32768 tokens.

Key Capabilities

  • Crystal Structure Generation: Generates crystal structures directly from natural language instructions, such as chemical formula and bulk modulus.
  • Reasoning Integration: Utilizes "thinking traces" to incorporate crystallographic and physical priors before generating atomic coordinates.
  • Reinforcement Learning (RL): Employs RL with verifiable rewards to enhance the validity, stability, and property conditioning of the generated structures.
  • Pymatgen Compatibility: Provides a script to convert generated text output into pymatgen.core.Structure format for further analysis and use.

Good For

  • Researchers and engineers in materials science and chemistry who need to generate hypothetical crystal structures.
  • Automating the initial design phase of new materials based on desired properties.
  • Exploring the space of possible crystal structures under specific constraints.