SL-AI/GRaPE-2.5-Quasar

VISIONPricing:Input $1.6 / Cached $0.15 / Output $12Concurrent Unit Cost:2Model Size:27BQuant:FP8Context Size:32kTool Calling:SupportedPublished:Sep 7, 2026License:apache-2.0Architecture:Transformer0.0K Open Weights Featherless Exclusive Cold

SL-AI/GRaPE-2.5-Quasar is a 27 billion parameter multimodal language model from SLAI (Skinnertopia Lab for Artificial Intelligence), supporting image and text inputs to produce text outputs. It features controllable thinking modes and work efforts for agentic tasks, with a focus on improved data quality and a less robotic prose style. This model excels in creative tasks, agentic coding, and STEAM domains, offering enhanced generalization and a unique approach to idea generation.

Loading preview...

GRaPE 2.5 Quasar: Multimodal Agentic Reasoning

GRaPE 2.5 Quasar is the flagship 27 billion parameter model from SLAI's third-generation GRaPE family, designed for multimodal inputs (image + text) and text outputs. It builds upon GRaPE 2 Pro with substantially improved training data quality and a focus on eliminating common "LLM-isms" for a more natural prose.

Key Capabilities & Features

  • Multimodal Input: Accepts image, video, and text inputs.
  • Controllable Reasoning: Features <thinking_mode=xxx> tags (minimal, low, medium, high, xtra-hi) to adjust reasoning depth, optimizing token usage for agentic tasks.
  • Controllable Work Effort: Utilizes <work_effort=xxx> tags (minimal, low, medium, high, xtra-hi) to manage tool usage and analysis depth for agentic tasks.
  • Enhanced Creativity: Trained with a significant focus on creative conversations and storytelling, aiming for unique and imaginative outputs.
  • Agentic Coding: Improved performance in agentic coding tasks by preserving previous thinking blocks in context.
  • Idea Generation: Scaffolds entire thinking blocks with a TODO-list style and generates multiple ideas, weighing approaches to minimize off-task deviations.
  • Multilingual Thinking: Capable of thinking in languages other than English, with seamless language adaptation.

Use Cases

GRaPE 2.5 Quasar is particularly well-suited for:

  • Creative Tasks: Generating unique stories, conversations, and imaginative content.
  • Agentic Coding: Complex coding tasks requiring multi-step reasoning and tool utilization.
  • STEAM Domains: Applications in Science, Technology, Engineering, Arts, and Mathematics.
  • Deep Analytical Work: Tasks requiring extended and deep thought processes.

Finetunes
2 models