Mistral AI's Ministral-3-3B-Instruct-2512 and Ministral-3-8B-Instruct-2512 models are now accessible via Amazon SageMaker JumpStart. These models are designed for edge deployment and resource-limited environments, offering a compact yet powerful AI solution. The Ministral-3-3B model is optimized for ultra-lightweight edge use cases with multimodal understanding. It features a 3.4B language model and a 0.4B vision encoder, fitting within 8GB of VRAM. This model supports a 256K-token context window and excels in vision analysis and multilingual instructions. It also demonstrates strong system prompt adherence and native function calling capabilities. The Ministral-3-8B model provides frontier-class performance in a compact 8B package. It includes an 8.4B language model and a 0.4B vision encoder, requiring 12GB of VRAM. This model utilizes interleaved sliding-window attention for efficient inference. Both models offer impressive multilingual support, agentic capabilities, and function calling. Customers can easily deploy these models through SageMaker JumpStart for various AI applications.