AuroraSystem/Clary-0.6-0.6B

TEXT GENERATIONConcurrent Unit Cost:1Model Size:0.8BQuant:BF16Context Size:32kTool Calling:SupportedPublished:Aug 13, 2026License:apache-2.0Architecture:Transformer0.0K Open Weights Featherless Exclusive Cold

AuroraSystem/Clary-0.6-0.6B is a 0.8 billion parameter multimodal language model developed by AuroraSystem, built upon the Qwen3-0.6B architecture and enhanced with a CLIP ViT-B/32 vision encoder and a custom projector. This model supports a 32768 token context length and excels in both Russian and English text processing, Python code generation, mathematics, and multimodal vision tasks like image description. Its primary differentiator is its integrated vision capabilities within a compact 0.8B parameter footprint, making it suitable for multimodal applications where resource efficiency is key.

Loading preview...

AuroraSystem Clary-0.6-0.6B: A Compact Multimodal LLM

AuroraSystem/Clary-0.6-0.6B is a 0.8 billion parameter multimodal language model based on the Qwen3-0.6B architecture. It integrates a frozen CLIP ViT-B/32 vision encoder and a custom vision projector (768→2048→2048→1024) to process image inputs, converting them into 49 vision tokens that are then combined with text tokens for the Qwen3-0.6B base model. This architecture enables the model to handle both textual and visual information efficiently.

Key Capabilities

  • Multilingual Text Processing: Supports instructions, knowledge retrieval, and summarization in both Russian and English.
  • Code Generation: Proficient in generating Python code.
  • Mathematical Reasoning: Capable of solving mathematical problems.
  • Multimodal Vision: Can describe images and identify colors, with an integrated vision architecture.
  • Thinking Mode: Includes a /think mode, leveraging Qwen3's thinking capabilities.

Limitations

  • OCR Weakness: Struggles with recognizing small text in images.
  • System Prompts: Not specifically trained on system prompts.
  • Performance Ceiling: Its performance is inherently limited by its 0.6B parameter base model class.

Good for

  • Applications requiring a compact, multimodal model for image understanding and text generation.
  • Developers needing Python code generation and mathematical problem-solving in a small footprint.
  • Use cases involving Russian and English text summarization and instruction following.

This model is available in alternative formats like GGUF for compatibility with llama.cpp, LM Studio, and Ollama, and is licensed under Apache-2.0.

Finetunes
2 models