Mistral API Pricing in 2026: Large 4, Medium 3.5 and Small 4 Are Different SKUs

Read the model identity before the dollar sign

On the official USD Standard API table, Mistral Large 4 shows $1.36 input, $0.14 cached input and $4.18 output per million tokens. Mistral Medium 3.5 shows $1.50, $0.15 and $7.50, whereas Mistral Small 4 shows $0.15, $0.015 and $0.60. The fact that one flagship input rate is below a medium-family rate is not a mistake to “correct” by assumption: published rates reflect different commercial SKUs. Name the model in every ledger row.

Separate October launch promotion from the list rate

Mistral marked Large 4 public preview on 6 October 2026, announced fifty percent off launch pricing for two weeks, and listed the model identifier as mistral-large-4. The model-card standard table displays the un-discounted $1.36/$0.14/$4.18 prices. A historical promotional row would need its precise effective and expiration window; the release note gives a duration but not a precise timezone-specific boundary. We archive the standard rate without inventing its first effective minute.

How cached usage changes a worked budget

Consider one million non-cached input tokens, half a million eligible cached input tokens and 200,000 output tokens on Large 4. At the documented standard list rates, the illustrative cost is 1×1.36 + 0.5×0.14 + 0.2×4.18 = $2.266, before regional modifiers and any other product charges. Cache eligibility is not established by repeating text alone: measure the tokens the platform actually classifies as cached and bill the distinct counter.

Batch, priority and regional inference are separate offers

Mistral pricing exposes switches for Standard, Batch and Priority service and regional inference. A number read from the Standard USD table cannot automatically be labeled a Batch EUR quote. Mistral’s model pages show capability and release stage, but a developer must still verify endpoint, country, organization tier and billing currency. Compare different models through the same measured workload; do not compare list prices with an undisclosed private commitment rate.

Practical editorial decision

For high-volume categorization, Small 4 may be a cost candidate; for output-heavy multimodal and agentic work, Medium 3.5 and Large 4 need full end-to-end trials by the owning model-evaluation team. This pricing guide deliberately does not claim which is more accurate or faster. Record token usage, cache hit rates, batch eligibility and human review cost. Historical prices should be snapshotted rather than silently updated when a preview becomes generally available.

How to verify the actual invoice

Begin with a small representative workload rather than the cheapest published headline. Record the exact provider and billing channel, API model identifier, currency, account geography, plan and commitment length. The price or quota evidence in this guide is keyed to cached_input_tokens_per_1m, input_tokens_per_1m, output_tokens_per_1m. Do not combine proprietary creative credits, token meters or media seconds into a single currency or invent a conversion rate between them. If an enterprise or regional rate is not published, leave its price unknown and obtain a written quote.

Keep separate estimates for failed attempts, optional tools, storage, exports, human quality review and tax. Use source-verified line items from the companion CSV instead of typing permanent numbers into your forecasting formulas. Where an old plan or model was superseded, preserve its original historical record and create an additive later snapshot. This guide does not claim a product test, performance ranking, independent legal assessment or live procurement approval.

Availability, alternatives and editorial limits

Available plans, supported endpoints, trial eligibility, quota reset and contract rights may vary by region and account even when a public website displays a single global currency. Compare an alternative only with the same accepted task, input/output volume, model version and privacy requirements. Vendor-provided marketing claims are not independent evidence of quality. This historical pricing article may be cross-referenced by the model, media or enterprise procurement team, but its purpose is to explain the billing mechanics without duplicating their evaluations or inventing test scores.