Undi95/MistralThinker-v1.1

Hugging Face
TEXT GENERATIONPricing:Input $0.7 / Cached $0.04 / Output $1.16Concurrent Unit Cost:2Model Size:24BQuant:FP8Context Size:32kTool Calling:SupportedPublished:Feb 26, 2025Architecture:Transformer0.1K Featherless Exclusive Warm

MistralThinker-v1.1 by Undi95 is a 24 billion parameter model, distilled from Mistral-Small-24B-Base-2501 using a DeepSeek R1 process. It is primarily designed for roleplay (RP) and storywriting applications, with 40% of its training data focused on these creative tasks. The model excels at generating narratives, character interactions, and immersive storytelling, supporting a context length of 32768 tokens.

Loading preview...

MistralThinker-v1.1: Specialized for Roleplay and Storywriting

MistralThinker-v1.1, developed by Undi95, is a 24 billion parameter model derived from Mistral-Small-24B-Base-2501 through a DeepSeek R1 distillation process. This model is uniquely optimized for creative text generation, with approximately 40% of its extensive training dataset dedicated to roleplay, storywriting, and character card data. It inherits a refined chain-of-thought process from DeepSeek R1, emphasizing narrative coherence and immersive outputs.

Key Capabilities

  • Rich Storytelling & Roleplay: Excels in generating creative narratives, detailed character portrayals, and engaging scenario building.
  • Contextual Dialogue: Capable of sustaining deep, context-driven conversations and interactions.
  • Character Lore Generation: Ideal for crafting and expanding character backstories and relationships.
  • Flexible Usage: Can be effectively used with or without a system prompt, adapting to various user preferences for direct storytelling or roleplay scenarios.

Good For

  • Roleplay (RP) Enthusiasts: Engaging in fictional or scenario-based interactions with a highly responsive AI.
  • Writers & Creators: Generating narratives, dialogues, and creative texts for stories, games, or other projects.
  • Character Development: Crafting detailed character profiles and backstories.

While strong in creative generation, users should be aware of potential hallucinations in the thinking process, though these often lead to successful replies. The model's dataset size was doubled from previous versions, enhancing its creative and contextual understanding.

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
–