Forge is available: build a website with AI from your RodiumAi account.
Try Forge
Mistral family — open and commercial AI models focused on efficiency, multilingual capabilities, and enterprise deployment.
Available models
8
Starting rate
30.2 RODI / 1M
Max context
256K
Frontier open-weight MoE model (39B active / 673B total) on Microsoft Foundry. Multimodal chat with text and image inputs, long-context comprehension, tool calling, and strong instruction following. Apache 2.0. Global Standard deployment.
Price
Enterprise document understanding on Microsoft Foundry. Combines mistral-ocr-2512 OCR with layout-aware parsing: PDFs, scans, and images → structured Markdown (with interleaved figures), JSON extraction, tables, handwriting, and multilingual content. Optimized for invoices, contracts, forms, and high-volume OCR pipelines — not general chat.
Price
Agentic coding model on Amazon Bedrock (123B, FP8). Specialized for autonomous software engineering: multi-file edits, repository exploration, tool calling, and SWE-bench-class performance with a 256K context window.
Price
Mistral reasoning model (~24B) on Amazon Bedrock. Multimodal text+vision assistant tuned for logic, coding copilots, and cost-efficient extended thinking at 128K context.
Price
Largest Ministral 3 edge model on Amazon Bedrock. Dense 13.5B LM + vision encoder, 256K context, function calling, and multimodal chat optimized for single-GPU deployment.
Price
Mid-size Ministral 3 on Amazon Bedrock. Dense 8.4B LM + vision encoder with 256K context, function calling, and strong text/vision performance for edge and single-GPU setups.
Price
Smallest Ministral 3 on Amazon Bedrock. Compact 3.4B LM + vision encoder for low-latency edge deployment, multimodal chat, tools, and 256K context at minimal cost.
Price
Speech-centric Voxtral model on Amazon Bedrock. Compact 3B voice assistant for low-latency ASR, spoken-language understanding, and text responses via Converse (audio + text in, text out).
Price