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

KIEFERSA/Sophea-Titan-1 is a 27 billion parameter multimodal chat model developed by Kiefer SA, built upon Qwen3.6-27B. It is specifically fine-tuned for Greek conversational tasks, offering formal and informal register control while retaining strong English language and vision capabilities. This model excels as a general-purpose Greek conversational assistant and for Greek-knowledge QA, outperforming other open-weight models in its size class on general Greek benchmarks.

Loading preview...

Overview

Sophea-Titan-1 is a 27 billion parameter multimodal chat model from Kiefer SA, based on Qwen3.6-27B. It is uniquely fine-tuned for the Greek language, incorporating register control and a fixed assistant identity, while maintaining its original English proficiency and multimodal (vision) capabilities. A key feature is its "non-thinking" mode (enable_thinking=false), which is crucial for optimal Greek output quality.

Key Capabilities

  • Multilingual Proficiency: Primarily focused on Greek, with strong retention of English language abilities.
  • Multimodal: Supports vision inputs, allowing for image understanding and description tasks.
  • Register Control: Offers the ability to switch between formal and informal conversational styles in Greek.
  • Optimized Greek Performance: Achieves a macro score of 0.7369 on general Greek benchmarks, surpassing its base model and other comparable open-weight models. It also shows a vision score of 0.714 on MMStar.
  • Efficient Deployment: Available in various quantized formats (FP8, NVFP4, GGUF, MLX) for reduced VRAM usage, with minimal impact on performance.
  • Speculative Decoding: Includes a multi-token-prediction head for enhanced inference speed, compatible with vLLM versions 0.23.0 and above.

Good For

  • General-purpose Greek conversational AI applications.
  • Question answering systems requiring Greek language knowledge.
  • Multimodal tasks involving image understanding and text generation in Greek or English.
  • Applications requiring controlled conversational register (formal/informal) in Greek.

Finetunes
5 models