Madras1/Jade0.6b
TEXT GENERATIONConcurrency Cost:1Model Size:0.8BQuant:BF16Ctx Length:32kPublished:Mar 25, 2026License:apache-2.0Architecture:Transformer Open Weights Warm

Madras1/Jade0.6b is an 0.8 billion parameter conversational finetune of Qwen3, specifically designed for Brazilian Portuguese. This model is optimized to express a strong, persistent persona, utilizing colloquial language, abbreviations, and slang for PT-BR chat and chatbot use cases. Its primary strength lies in maintaining an informal, friendly, and chat-oriented tone, making it suitable for character-style interactions.

Loading preview...