EnlistedGhost/Magistral-Small-2509-Vision
EnlistedGhost/Magistral-Small-2509-Vision is a 24 billion parameter multimodal model re-configured by EnlistedGhost from MistralAI's Magistral-Small-2509, featuring re-enabled vision functionality. This model utilizes MistralAI's "Mistral3ForConditionalGeneration" processor and custom chat templates, supporting both vision and text inputs. It is designed for general inferencing tasks where both visual and textual understanding are required, building upon the original Magistral-Small-2509 architecture.
Loading preview...
Overview
EnlistedGhost/Magistral-Small-2509-Vision is a 24 billion parameter multimodal model derived from MistralAI's Magistral-Small-2509. Developed by EnlistedGhost, this version re-enables and integrates vision functionality, making it capable of processing both visual and textual inputs. It uses a re-configured "Mistral3ForConditionalGeneration" processor and includes custom chat templates and system prompts, differing from MistralAI's original configurations.
Key Capabilities
- Multimodal Processing: Supports both vision and text inputs, allowing for tasks requiring understanding across modalities.
- Custom Configuration: Features custom chat templates and system prompts for potentially optimized performance and usability.
- Ease of Use: Designed to be directly usable with Huggingface SafeTensors, Pipeline, or vLLM without additional modifications.
Important Considerations
- Evaluation Status: This model has not been formally evaluated in any capacity, and users should proceed with caution.
- Custom Prompts: While standard MistralAI templates can be used, the model is provided with custom chat templates and system prompts that EnlistedGhost found to be effective.
Intended Use
This model shares the same intended use cases as the original MistralAI Magistral-Small-2509, focusing on general language and now vision-enhanced tasks. Users should be aware of the lack of formal evaluation and use it at their own risk.