SL-AI/Openprose-2-Flash

VISIONConcurrent Unit Cost:1Model Size:9BQuant:FP8Context Size:32kTool Calling:SupportedPublished:Jul 19, 2026License:apache-2.0Architecture:Transformer0.0K Open Weights Featherless Exclusive Cold

SL-AI/Openprose-2-Flash is a 9 billion parameter multimodal language model built on a Qwen3.5 base, developed by SLAI (Skinnertopia Lab for Artificial Intelligence). It accepts image and text inputs to generate text outputs, specifically fine-tuned on 200,000 entries of human-text for creative writing tasks. This model excels in generating complex story ideas and is optimized for on-device creative writing applications.

Loading preview...

Openprose 2 Flash Overview

SL-AI/Openprose-2-Flash is a 9 billion parameter multimodal model developed by SLAI (Skinnertopia Lab for Artificial Intelligence), designed specifically for creative writing. Built upon a Qwen3.5 base, it significantly improves upon its predecessor, Openprose 1, by utilizing a stronger foundation model and undergoing more extensive training on creative writing data.

Key Capabilities & Features

  • Multimodal Input: Accepts both image and text as input, producing text as output.
  • Creative Writing Focus: Post-trained on a curated dataset with heavy emphasis on creative writing, making it highly specialized for generating stories and creative content.
  • Enhanced Base Model: Leverages the more capable Qwen3.5 9B foundation, offering improved performance over the previous generation's Qwen3 VL base.
  • Increased Parameter Count: With 9 billion parameters, it boasts a 6 billion parameter increase over Openprose 1, contributing to its enhanced ability to handle complex ideas in storytelling.
  • Open-Source Training Data: The training data used for Openprose 2 Flash is open-source.

Intended Use & Limitations

Openprose 2 Flash is specifically optimized for creative writing and will perform poorly on other domains. It does not support any "thinking" capabilities; its strength lies purely in text generation based on its specialized training. Recommended inference settings are provided for optimal performance, including a temperature of 0.6 and Top P of 0.95.