SL-AI/GRaPE-2-Pro

Hugging Face
VISIONConcurrent Unit Cost:2Model Size:27BQuant:FP8Context Size:32kTool Calling:SupportedPublished:Apr 19, 2026License:apache-2.0Architecture:Transformer0.0K Open Weights Featherless Exclusive Warm

SL-AI/GRaPE-2-Pro is a 27 billion parameter multimodal language model developed by Skinnertopia Lab for Artificial Intelligence (SLAI), built on a Qwen3.5 base. It accepts image and text inputs to produce text outputs, featuring an extended thinking mode system for controllable reasoning depth. This model is specifically post-trained with a heavy emphasis on code, STEAM subjects, and logical reasoning, making it suitable for complex problem-solving and structured analytical tasks.

Loading preview...

GRaPE 2 Pro: Flagship Reasoning Model

GRaPE 2 Pro is the flagship 27 billion parameter model from SLAI's second-generation GRaPE family, built upon a robust Qwen3.5 base. It is a multimodal model, processing both image and text inputs to generate text outputs. A key differentiator is its unique "thinking mode" system, allowing users to control the depth of reasoning from minimal to xtra-Hi via a prompt tag, optimizing for task complexity and inference speed.

Key Capabilities & Features

  • Multimodal Input: Processes both images and text.
  • Controllable Reasoning: Features six discrete thinking tiers (minimal, low, medium, high, xtra-Hi, auto) for adaptable problem-solving.
  • Specialized Training: Post-trained on a proprietary dataset with significant emphasis on:
    • Code (~50% of post-training data)
    • STEAM (Science, Technology, Engineering, Arts, Mathematics)
    • Logical reasoning and structured problem solving
  • Stronger Base: Utilizes Qwen3.5-27 as its foundation, enhancing overall performance.

Ideal Use Cases

GRaPE 2 Pro is particularly well-suited for applications requiring:

  • Complex Code Generation: Benefits from extensive code training and deep reasoning modes.
  • Multi-step Mathematical Problems: Leverages its logical reasoning capabilities.
  • Deep Analytical Work: The high and xtra-Hi thinking modes are designed for intricate analysis.
  • Agentic Workflows: Low or Auto thinking modes are recommended for faster actions in agent-based systems.

This model aims to provide strong structured reasoning capabilities while remaining deployable on consumer hardware.

Finetunes
2 models