KIEFERSA/Sophea-Titan-1
KIEFERSA/Sophea-Titan-1 is a 27 billion parameter multimodal chat model developed by Kiefer SA, fine-tuned from Qwen3.6-27B. It specializes in Greek conversational AI with register control and a fixed assistant identity, while retaining strong English language and vision capabilities. This model achieves a general Greek benchmark score of 0.7369 and an English retention score of 0.8783, making it suitable for general-purpose Greek chat and knowledge-based QA.
Loading preview...
Overview
Sophea-Titan-1 is a 27 billion parameter multimodal chat model developed by Kiefer SA, built upon the Qwen3.6-27B architecture. It 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 from its base model.
Key Capabilities
- Primary Language: Optimized for Greek, offering general-purpose conversational assistance and knowledge-based QA.
- Multimodal: Retains vision capabilities, allowing for image understanding and interaction.
- Register Control: Supports formal and informal conversational registers in Greek.
- Performance: Achieves a macro score of 0.7369 on general Greek benchmarks and 0.8783 for English retention, outperforming its base model and other comparable open-source Greek LLMs.
- Non-Thinking Mode: Designed to operate optimally with
enable_thinking=falsein its chat template, which is crucial for maintaining Greek output quality. - Speculative Decoding: Includes a multi-token-prediction head for improved inference throughput with vLLM (version 0.23.0 or newer).
Good For
- Applications requiring a robust Greek conversational AI.
- Greek-language knowledge question-answering systems.
- Multimodal applications where Greek language processing and image understanding are needed.
- Developers looking for a performant Greek model that also retains strong English capabilities.