Dynamically estimate Mistral API billing spanning Mistral Large 3, Devstral coding agents, Voxtral voice streaming metrics, and Native Batching.
Add native Tool-use invocations (e.g. Agentic Search, OCR Pages, or internal code logic calls). Priced universally:
Flagship multimodal and multilingual model, open-weight and general-purpose.
New standard for multimodal, reasoning-optimized models with agentic capabilities.
State-of-the-art performance for simplified enterprise deployments and cost-efficiency.
The world’s best document extraction and understanding model with multimodal vision.
Enhanced model for advanced coding agents and autonomous software engineering.
The best open-source model optimized specifically for lightweight coding agents.
Lightweight, fast model proficient in over 80 programming languages.
The first open-source code agent specifically designed for the Lean 4 language.
Multimodal and multilingual SOTA model licensed under Apache 2.0.
Reasoning model excelling in domain-specific, transparent reasoning tasks.
Lightweight reasoning model offering transparent and multilingual understanding.
Family of models (3B, 8B, 14B) designed for frontier AI performance on the edge.
State-of-the-art text-to-speech engine for natural voice generation.
Instant voice clarity and transcription built for low-latency edge use cases.
Speech-to-text model with multimodal audio and text understanding.
A specialized classifier service for precise text content moderation.
Vision-capable large model with frontier reasoning capabilities.
Small-form state-of-the-art model trained specifically for code tasks.
The most performant open model currently available, using Mixture-of-Experts.
Mistral’s Agent API enhances AI workflows with built-in tools for code execution, web search, image generation, and persistent library memory. These specialized tools are billed per invocation (e.g., $30 per 1,000 web searches) additionally to standard token context processing.
Enterprise APIs introduce regional data processing controls and system-level SLAs. Mistral offers premium news access ($50/1k calls) and data capture features ($0.04/M tokens) to allow for continuous debugging and optimization in sensitive enterprise environments.
OCR 3 is specialized for complex document understanding. It charges $2 per 1,000 pages for standard extraction and $3 per 1,000 pages for page annotations. Batch API users can save an additional 50% on these document processing tasks.
The Voxtral family supports everything from text-to-speech ($0.016 per 1k characters) to real-time transcription. Multimodal variants like Voxtral Small allow you to process audio and text simultaneously in a single chat completion turn natively.
Technical details based on Mistral's latest 2026 API documentation.
You can save 50% on all token costs (input and output) by using Mistral’s Batch API for non-time-sensitive requests. Requests are typically processed within 24 hours. Note that while OCR, text, and reasoning models support batch discounts, image generation is billed at standard rates.
Mistral Large 3 supports an expansive context window of 128,000 tokens as a standard, with higher limits available for Enterprise users. Open-weight versions like Mixtral 8x22B also support massive context handling natively.
Mistral charges a $0.05 fee per request for violations caught before generation in the Responses API. If a violation is deemed to have occurred during generation, the standard token costs for that generation still apply.
Codestral is optimized for performance across over 80 programming languages, making it one of the most versatile coding models for both legacy and modern software stacks.
Enterprise APIs with regional controls and premium support are customized based on volume and requirements. They often include higher rate limits and guaranteed system-level SLAs for mission-critical deployments.
Yes, for fine-tuned models like Codestral or Ministral variants, Mistral charges a flat $2 per month per model for storage, in addition to the initial training costs (which range from $1 to $4 per Million tokens).