This repository contains GGUF format model files forQwen’s Qwen3.5-0.8B. Qwen3.5-0.8B is a highly compact yet capable model featuring a Hybrid Gated DeltaNet + Gated Attention architecture and a unified vision-language foundation. Despite having only 0.8 billion parameters, it natively supports up to 262,144 tokens of context, tool calling, and multimodal inputs (images/video). > Note: Due to its parameter scale, this model operates in non-thinking mode by default and is best suited for prototyping, task-specific fine-tuning, or lightweight edge deployments. To see the available files, please verify the Files and versions tab. Important: This model uses a hybrid Gated DeltaNet architecture. Ensure you are using the absolute latest version of llama.cpp to support these new operators. Recommended Parameters: Qwen3.5-0.8B operates in Non-Thinking Mode by default. Temperature: 1.0 (Text tasks) or 0.7 (Vision/Language tasks) Top-P: 1.0 (Text) or 0.8 (Vision) Context: -c` (The model supports ultra-long context, adjust based on your VRAM/RAM). > [!Note] > This repository contains model weights and configuration files for the post-trained model in the Hugging Face Transformers…
Модальности:
Генерация текста Компьютерное зрение Мультимодальность
Области применения:
Диалог / чат
Задача: Генерация текста
Автор: AaryanK
Теги: gguf, text-generation-inference, qwen, multimodal, vision, agent, en, endpoints_compatible
Лайков: 4 | Загрузок: 327
Описание основано на материалах HuggingFace. Перевод выполнен автоматически.
