TheDrummer/Gemma-3-R1-27B-v1

Hugging Face
VISIONConcurrency Cost:2Model Size:27BQuant:FP8Ctx Length:32kPublished:Aug 4, 2025Architecture:Transformer0.0K Warm

TheDrummer/Gemma-3-R1-27B-v1 is a 27 billion parameter Gemma 3 series model developed by TheDrummer, fine-tuned for enhanced reasoning capabilities. This model is designed to unlock more advanced reasoning and creative roleplay, potentially offering vision capabilities. It excels in generating creative and less positive responses, making it suitable for complex narrative and interactive applications.

Loading preview...

TheDrummer/Gemma-3-R1-27B-v1 Overview

This model is a 27 billion parameter variant from the Gemma 3 series, developed by TheDrummer. It is specifically fine-tuned to enhance reasoning capabilities and is noted for its creative output and less overtly positive responses. The model is also described as potentially vision-capable.

Key Capabilities

  • Enhanced Reasoning: The model is a "reasoning tune" designed to unlock more advanced cognitive functions.
  • Creative Roleplay: Users report it to be highly creative in roleplay scenarios, with a non-sloppy writing style.
  • System Prompt Adherence: Demonstrates rigid adherence to system prompts, similar to Cydonia R1 24B.
  • Vision Capability: The model is indicated to be vision-capable, suggesting multimodal potential.
  • Unique Thinking Tags: Utilizes a <think> prefill token for assistant turns, allowing for creative modifications like <evil_think> or <creative_think> to guide reasoning.

Good For

  • Complex Roleplay: Excels in multi-character roleplay and creative narrative generation.
  • Reasoning-Intensive Tasks: Suitable for applications requiring deeper logical processing and structured thought.
  • Customizable AI Personalities: The flexible thinking tags allow for nuanced control over the model's internal monologue and response style.

Popular Sampler Settings

Top 3 parameter combinations used by Featherless users for this model. Click a tab to see each config.

temperature
top_p
top_k
frequency_penalty
presence_penalty
repetition_penalty
min_p