KIEFERSA/Sophea-Titan-1

VISIONPricing:Input $1.6 / Cached $0.15 / Output $12Concurrent Unit Cost:2Model Size:27BQuant:FP8Context Size:32kTool Calling:SupportedPublished:Jul 12, 2026License:apache-2.0Architecture:Transformer0.0K Open Weights Featherless Exclusive Cold

Sophea-Titan-1 is a 27 billion parameter multimodal chat model developed by Kiefer SA, fine-tuned from Qwen3.6-27B. It is primarily optimized for Greek conversational tasks, offering register control and a fixed assistant identity, while retaining strong English language and vision capabilities. The model achieves a general Greek benchmark macro score of 0.7369 and an English retention macro score of 0.8783, outperforming its base model and other open-source Greek LLMs in its size class. Its key differentiator is its "non-thinking" decoding mode, which is crucial for maintaining Greek output quality.

Loading preview...

Sophea-Titan-1: A Greek-Optimized Multimodal Chat Model

Sophea-Titan-1 is a 27 billion parameter multimodal chat model developed by Kiefer SA, built upon the Qwen3.6-27B architecture. This model is specifically fine-tuned for Greek conversational interactions, incorporating register control and a consistent assistant identity. Notably, it maintains strong English language proficiency and multimodal vision capabilities.

Key Capabilities

  • Primary Greek Language Support: Optimized for general-purpose Greek conversational assistance and Greek-knowledge QA.
  • Multimodal Vision: Retains and enhances vision capabilities from its base model, achieving an MMStar score of 0.714.
  • English Language Retention: Demonstrates strong English performance, with an English retention macro score of 0.8783.
  • Register Control: Offers formal and informal register control for nuanced Greek conversations.
  • "Non-Thinking" Decoding: Features a unique decoding mode (enable_thinking=false) that is essential for optimal Greek output quality.

Good For

  • Applications requiring a high-performance Greek conversational AI.
  • Multilingual chat assistants where both Greek and English are critical.
  • Vision-enabled tasks within a Greek language context.
  • Developers seeking a specialized Greek model that outperforms other open-source alternatives in its size category, as evidenced by its general Greek benchmark macro score of 0.7369.

Finetunes
5 models