Mistral API: European Frontier Models

Overview

Mistral AI is a Paris-based lab producing frontier language models with a focus on efficiency and European data sovereignty. Its API follows the OpenAI SDK format, and Mistral models are widely available on third-party inference providers including Together AI, Groq, and OpenRouter.

Pricing (May 2026)

Model	Input $/M	Output $/M
Mistral Large 3	$2.00	$6.00
Mistral Medium 3	$0.40	$2.00
Mistral Small 3.1	$0.20	$0.60
Mistral Small 4	$0.15	varies
Ministral 8B	$0.10	$0.10
Mistral Nemo	$0.02	varies

Mistral Large 3 output at $6/M is 40% below GPT-5.4 and 60% below Claude Sonnet 4.6 -- a meaningful cost advantage for output-heavy workloads like document generation and long-form code writing.

A free "Experiment" tier is available with rate limits, no credit card required, suitable for prototyping.

What distinguishes Mistral

Mistral models are developed and trained in Europe, which matters for teams with GDPR or EU AI Act compliance requirements around model provenance. Mistral operates its own inference infrastructure and offers data residency in the EU.

Le Chat Pro ($14.99/mo) is the consumer product -- it does not include API credits. The API and the consumer subscription are billed independently.

Where it fits

Mistral Large 3 is the target model for teams where GPT-5.4 pricing is painful and Claude Sonnet's capabilities are more than required. The output cost advantage is most visible in batch pipelines that generate large volumes of text.

For teams that need flexibility across Mistral and other providers, OpenRouter carries the full Mistral model catalog alongside 280+ others.

Field notes

• Mistral Small 4 pricing confirmed at $0.15/M input in the May 2026 pricing page update -- the cheapest multimodal-capable model in the Mistral lineup. Teams using Mistral Small 3.1 for vision tasks should evaluate the upgrade path. [changelog, 2026-05-01]

See also

OpenAI API, Anthropic API, Together AI, OpenRouter