Mistral Medium 3.5

The Mistral Medium 3.5 family from Mistral AI introduces flagship dense multimodal model that unify instruction following, reasoning, and coding within a single architecture. It combines configurable reasoning, native vision capabilities, multilingual support, and long-context performance for versatile agentic applications.

1 variant4 quantized variants

Mistral Medium 3.5 128B

Mistral

Params: 398B

Mistral Medium 3.5 is a 128B-parameter dense multimodal language model designed for instruction following, reasoning, coding, and agentic workflows. It features 88 transformer layers, 96 attention heads, and a 256K-token context window, while accepting text and image inputs with text output. The model supports configurable reasoning effort, native function calling, structured JSON generation, multilingual capabilities, and a vision encoder trained from scratch to process variable image sizes and aspect ratios.

Text GenerationInstruction Following+8

Runs onB200+

Runs on

Mistral Medium 3.5 Models | Vultr Docs