benchrPricing breakdown
| Tier | Rate / 1M tokens |
|---|---|
| Standard input | $1.50 |
| Standard output | $7.50 |
| Context window | 128,000 tokens |
Vision and PDF: what it enables
Mistral Medium 3.5 processes images and PDF documents natively in the same API call as text. Practical applications: invoice and receipt extraction (image + OCR in one step); contract review where PDFs contain embedded charts and tables; form processing with signature image detection; product catalog analysis combining images and specifications. The PDF support is particularly valuable for EU enterprise use cases — financial statements, regulatory filings, legal contracts — where documents frequently contain mixed text and figure content that pure text models cannot process.
Mistral Medium vs Mistral Large: the output cost consideration
Both Mistral Medium 3.5 and Mistral Large 3 are priced at $1.50/1M input, but their output rates differ significantly: Mistral Medium 3.5 output at $7.50/1M vs Mistral Large 3 at $1.50/1M — a 5× output price difference. For input-heavy tasks like document analysis where outputs are short summaries, Mistral Medium 3.5 is the right choice because you gain vision capability at the same input cost. For output-heavy tasks generating long reports or responses, Mistral Large 3's $1.50/1M output rate is dramatically cheaper. Evaluate your input-to-output ratio before choosing between the two.
EU residency for document processing
Document intelligence use cases often process sensitive content — contracts, medical records, financial documents, HR files. For EU organizations, processing this content through US-based APIs requires GDPR Article 46 transfer mechanisms (SCCs or BCRs). Mistral Medium 3.5 on La Plateforme processes document inputs on EU infrastructure, eliminating this transfer requirement. For legal, healthcare, and financial sector document pipelines, the compliance simplification is a meaningful operational benefit.
Cost scenarios
At 5M input + 1M output per month (document processing, short extraction outputs): Mistral Medium 3.5 costs $7.50 + $7.50 = $15/month. Gemini 3.1 Pro at the same volume: $10 + $12 = $22/month — Mistral is 32% cheaper for light document workloads. At 10M input + 5M output: Medium costs $15 + $37.50 = $52.50/month. The output cost ($7.50/1M) makes Mistral Medium 3.5 less competitive for generation-heavy workloads.
Use-case fit
Best for: EU document intelligence pipelines (invoices, contracts, regulatory filings); image-plus-text analysis where EU data residency is required; multimodal workloads on existing Mistral API integrations; GDPR-sensitive document processing without US data transfer complexity.
Skip if: You need video or audio support — Gemini 3.1 Pro or 3.5 Flash covers these. Also skip for output-heavy generation tasks; Mistral Large 3 ($0.50/$1.50) is dramatically cheaper on output per output token despite the same input price.
Decision checklist
Compare Mistral Medium 3.5 against Gemini 3.5 Flash at the same input price: if EU residency is required, Mistral is the cleaner path. If speed (289 tok/s on Gemini) or Google Search grounding matters more, Gemini 3.5 Flash is the better fit despite the US data residency requirement.
Check your output-to-input ratio: if outputs are 3+ tokens per input token, Mistral Medium 3.5's $7.50/1M output rate makes Mistral Large 3 (same input, 5× cheaper output) a better choice unless vision capability is strictly required.