OpenDFM/ChemDFM-v2.0-14B

Hugging Face
TEXT GENERATIONConcurrent Unit Cost:1Model Size:14.8BQuant:FP8Context Size:32kTool Calling:SupportedPublished:Oct 26, 2025License:agpl-3.0Architecture:Transformer0.0K Open Weights Featherless Exclusive Warm

ChemDFM-v2.0-14B is a 14.8 billion parameter dialogue foundation model developed by OpenDFM, specifically designed for Chemistry and molecule science. Built upon Qwen2.5-14B, it features upgraded domain pre-training with web-scale molecular and reaction data, including functional-group information, and enhanced instruction tuning for diverse chemical tasks. This model excels at acquiring fine-grained chemical knowledge and is optimized for chemistry-related dialogue and information retrieval.

Loading preview...

ChemDFM-v2.0-14B Overview

ChemDFM-v2.0-14B is the latest iteration of OpenDFM's pioneering open-source dialogue foundation model tailored for Chemistry and molecule science. This 14.8 billion parameter model is built on the Qwen2.5-14B architecture, significantly enhancing its chemical capabilities through a two-pronged approach.

Key Capabilities & Enhancements

  • Domain Pre-training: The model incorporates web-scale molecular and reaction data, enriched with functional-group information and properties. This allows ChemDFM-v2.0 to acquire chemical knowledge with finer granularity than previous versions.
  • Instruction Tuning: The instruction tuning dataset has been substantially improved, introducing a wider variety of tasks and increasing the diversity in phrasing and expression of instruction texts. This leads to more robust and versatile performance in chemistry-related dialogues.
  • Chemical Reasoning: While the README states it's a "non-thinking model," its specialized training aims to provide advanced general LLM capabilities within the chemistry domain.

Use Cases

ChemDFM-v2.0-14B is particularly well-suited for applications requiring detailed chemical descriptions, understanding of molecular properties, and engaging in chemistry-focused dialogue. Developers can leverage its specialized knowledge for tasks such as:

  • Generating detailed descriptions of molecules from SMILES notation.
  • Answering chemistry-specific questions.
  • Assisting in chemical research and education by providing structured information.

Users should note that while highly specialized, the model may still generate incorrect or misleading information, and verification with domain experts is recommended.

Finetunes
1 models