beezu/Magistry-24B-v1.1-mlx-bf16

Hugging Face
TEXT GENERATIONConcurrency Cost:2Model Size:24BQuant:FP8Ctx Length:32kPublished:Mar 22, 2026License:apache-2.0Architecture:Transformer Open Weights Warm

Magistry-24B-v1.1-mlx-bf16 is a 24 billion parameter language model, converted to MLX format from sophosympatheia/Magistry-24B-v1.1, which is a merge of several pre-trained models. Developed by sophosympatheia, this model is designed for creative tasks, aiming for enhanced coherency and entertaining output. It offers a 32768 token context length and is optimized for creative generation and conversational scenarios, with specific sampler settings provided for balancing flair and accuracy.

Loading preview...

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

Magistry-24B-v1.1 is a 24 billion parameter language model developed by sophosympatheia, converted to MLX format. It is an iteration on the Magistry-24B-v1.0 line, specifically engineered to improve coherency and handling while maintaining its creative and entertaining output. The model was created using the DELLA merge method, with Darkhn/Magistral-2509-24B-Text-Only as its base, and incorporates Casual-Autopsy/Maginum-Cydoms-24B, DarkArtsForge/Magistaroth-24B-v1, and huihui-ai/Huihui-Devstral-Small-2-24B-Instruct-2512-abliterated.

Key Capabilities & Features

  • Enhanced Creativity: Designed to produce entertaining and creative text, building upon its predecessor's "pizzazz."
  • Improved Coherency: Aims to offer better logical and physical continuity compared to v1.0, though it may still struggle with intricate details.
  • "Thinking" Mechanism: Supports explicit internal monologue via <think> tags to aid in planning and continuity for complex tasks.
  • Optimized Sampler Settings: Provides recommended sampler configurations (Conservative, Wild, Balanced) to fine-tune output for either accuracy or creative flair.
  • SillyTavern Integration: Includes a master import JSON for easy setup of system prompts, chat templates, and sampler settings in SillyTavern.

Use Cases & Considerations

  • Creative Writing & Roleplay: Excels in scenarios requiring imaginative and engaging text generation.
  • Conversational AI: Suitable for interactive applications where entertaining and dynamic responses are valued.
  • Handling Complex Scenarios: While improved, users may need to guide the model with explicit instructions or rerolls for highly detailed or logically demanding tasks.
  • Experimentation with Samplers: Users are encouraged to experiment with provided sampler settings to achieve desired output characteristics, balancing stability and creativity.

Finetunes
1 models