C3DS/CARDS-Qwen3.5-27B

VISIONConcurrency Cost:2Model Size:27BQuant:FP8Ctx Length:32kTool Calling:SupportedPublished:Apr 8, 2026License:apache-2.0Architecture:Transformer Open Weights Cold

C3DS/CARDS-Qwen3.5-27B is a 27 billion parameter language model fine-tuned by C3DS from the Qwen3.5-27B base model. It specializes in the classification of climate-contrarian claims using the CARDS taxonomy from Coan et al. (2025). This model demonstrates improved performance over its base model, achieving a Macro F1 of 0.766 on the CARDS test set, and is competitive with larger models like Claude Opus 4.6/4.7 for this specific task. It also retains the multimodal capability of the Qwen3.5 family, allowing for classification of claims from both text and image inputs.

Loading preview...

CARDS-Qwen3.5-27B: Climate-Contrarian Claim Classification

CARDS-Qwen3.5-27B is a 27 billion parameter model, fine-tuned from the Qwen3.5-27B base, specifically for classifying climate-contrarian claims. It utilizes the CARDS taxonomy developed by Coan et al. (2025), making it a specialized tool for analyzing climate discourse.

Key Capabilities

  • Specialized Classification: Excels at identifying and categorizing climate-contrarian claims based on the CARDS taxonomy.
  • Enhanced Performance: Outperforms the base Qwen3.5-27B model across all evaluation metrics on the CARDS test set, notably improving Macro F1 to 0.766 and eliminating parse failures.
  • Competitive Accuracy: Achieves a leading Macro F1 score among compared models, including Claude Opus 4.6/4.7, for rare-label coverage.
  • Multimodal Support: Inherits and preserves the base Qwen3.5 family's ability to process and classify claims from both text and image inputs.
  • Self-Contained Prompts: Includes bundled system prompts and CoT triggers with the CARDS taxonomy inlined for straightforward usage.

Good For

  • Researchers and analysts studying climate discourse and misinformation.
  • Applications requiring precise classification of climate-related claims.
  • Projects needing to analyze both textual and visual content for climate contrarianism.
  • Users seeking a specialized model with strong performance on a niche classification task, competitive with larger, general-purpose LLMs.

Finetunes
3 models