Mistral Large 3 pricing: 256K context and Apache 2.0 open weights

Mistral lists Mistral Large 3 at $0.50/1M input and $1.50/1M output, with a 256K-token context window, multimodal support, and open weights under Apache 2.0. Hosted data is in the EU by default, subject to endpoint and feature-specific caveats.

By benchr Editorial Team · · Pricing, context, license, multimodality, and data-location notes checked against official Mistral sources, July 24, 2026 · View changelog

Input / 1MMistral AI
Output / 1MMistral AI
Contextmax window
Languagesprovider reported

Pricing breakdown

mistral-large-3 — official Mistral AI pricing
TierRate / 1M tokens
Standard input$0.50
Standard output$1.50
Context window256,000 tokens

Apache 2.0 license: useful permissions with real terms

Mistral releases the base and instruction-tuned Large 3 weights under Apache 2.0. The license generally permits use, modification, and distribution, including commercial use, while retaining conditions around notices and attribution and separate rules for trademarks and patents. It is not a blanket promise of zero legal risk; read the license and apply it to your distribution and deployment model.

Hosted-data location: EU by default, with caveats

Mistral says hosted data is in the European Union by default and that explicitly using its US API endpoint stores data in the United States. It also says some features can temporarily transfer data outside the EU through listed subprocessors, with safeguards such as Standard Contractual Clauses where applicable. Region choice is relevant to governance, but it does not automatically establish GDPR, residency, or sector-specific compliance; review the current DPA, subprocessors, retention controls, and enabled features.

Multilingual performance

Mistral describes the Mistral 3 family as supporting 40+ native languages and reports strong multilingual results for Large 3. Treat that as provider-reported evidence, not a universal ranking: test the exact languages, dialects, document types, and quality criteria your workload requires.

Cost scenarios

At 20M input + 5M output per month, Mistral Large 3's published rates produce $10 + $7.50 = $17.50/month. A price-only comparison may put other APIs above or below that total, but price does not prove quality, legal suitability, or residency. Include task evaluation, hosting, tools, storage, taxes, and negotiated terms in a production estimate.

Use-case fit

Candidate for: General-purpose text-and-image workloads needing a 256K context window; teams evaluating Apache 2.0 self-hosting; multilingual applications; or hosted deployments where Mistral's regional controls fit the governance plan.

Compare alternatives if: Your workload requires a different price profile, a managed-service certification or region Mistral does not document for your plan, or stronger measured results on your specific coding, reasoning, or language tasks.

Decision checklist

Confirm the actual API endpoint and every enabled feature. If location is a requirement, document the DPA, subprocessor list, retention settings, and any feature-dependent transfer rather than relying on the provider's headquarters.

For self-hosting, review Apache 2.0's current text, preserve required notices, and assess trademarks, patents, downstream distribution, and third-party components with appropriate counsel.

Frequently asked

What license and context window does Mistral Large 3 use?

Mistral releases Large 3's base and instruction-tuned weights under Apache 2.0, and its current model card lists a 256K-token context window. Review the full license for notice, attribution, patent, trademark, and distribution considerations.

Where does Mistral host API data?

Mistral says hosted data is in the EU by default, while its explicit US endpoint stores data in the US. Some features can temporarily transfer data through subprocessors, so check the selected endpoint, features, DPA, retention settings, and current subprocessor list.

Is Mistral Large 3 a multimodal model?

Yes. Mistral's current model card describes Large 3 as a general-purpose multimodal model, and its vision documentation lists mistral-large-2512 for image input. Validate the supported formats and task quality for your workflow.

Changelog

  • — Corrected context to 256K and the API ID to mistral-large-2512; documented multimodality; qualified Apache, data-location, and compliance language.
  • — Expanded with license analysis, data-location guidance, multilingual context, and cost scenarios.
  • — Published. Pricing verified at mistral.ai/pricing.

Sources