sophosympatheia/Magistry-24B-v1.1

Hugging Face
TEXT GENERATIONConcurrency Cost:2Model Size:24BQuant:FP8Ctx Length:32kPublished:Mar 21, 2026License:apache-2.0Architecture:Transformer0.0K Open Weights Warm

Magistry-24B-v1.1 is a 24 billion parameter merged language model developed by sophosympatheia, built upon Darkhn/Magistral-2509-24B-Text-Only using the DELLA merge method. This model is designed to enhance coherency and handling while maintaining creative and entertaining output, making it suitable for creative text generation and interactive applications. It features a 32768 token context length and aims to improve logical and physical continuity over its predecessor.

Loading preview...

Magistry-24B-v1.1: A Creative Merge Model

Magistry-24B-v1.1 is a 24 billion parameter language model developed by sophosympatheia, representing an iteration designed to improve coherency and handling from its v1.0 predecessor without sacrificing its creative and entertaining output. Built using the DELLA merge method, its base model is Darkhn/Magistral-2509-24B-Text-Only, incorporating elements from Casual-Autopsy/Maginum-Cydoms-24B, DarkArtsForge/Magistaroth-24B-v1, and a pre-processed huihui-ai/Huihui-Devstral-Small-2-24B-Instruct-2512-abliterated model.

Key Capabilities & Characteristics

  • Enhanced Coherency: Aims to provide better logical and physical continuity compared to v1.0, though it may still struggle with intricate details.
  • Creative Output: Designed to be highly creative and entertaining, with specific sampler settings provided to optimize for either accuracy or flair.
  • Thinking Tags Support: Can utilize <think></think> tags for internal monologue and planning, which can assist in improving continuity and complex task handling.
  • Flexible Sampler Settings: Offers recommended "Conservative," "Wild," and "Balanced" sampler configurations, including Adaptive-P settings, to tailor output for different creative or accuracy-focused use cases.

Ideal Use Cases

  • Creative Text Generation: Excels in scenarios requiring imaginative and engaging content, such as role-playing or story generation.
  • Interactive Applications: Suitable for applications where entertaining and dynamic responses are prioritized.
  • Experimentation with Samplers: Provides detailed guidance and a master import JSON for SillyTavern users to experiment with various sampler settings to fine-tune model behavior.

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