Mistral AI Pricing 2026: Vibe (Le Chat) Pro, Team and API Costs per Token

Short answer: Mistral sells two separate things. The Vibe app (formerly Le Chat) costs $14.99 a month on Pro and $24.99 per user a month on Team, with a free plan and a $5.99 student price. The API bills per million tokens, from $0.10 on Ministral 3 3B to $1.50 input and $7.50 output on Mistral Medium 3.5. Checked on Mistral's pricing page and docs, 30 October 2026.

The mistake most buyers make is treating these as one bill. A Pro subscription does not turn into an API account for your app, and an API key does not give you the Vibe app. Pick the side you need first, then the plan or model.

Mistral AI pricing at a glance

App prices from Mistral's pricing page, API rates from Mistral's API pricing docs, standard tier, read on 30 October 2026. App prices exclude taxes.

What happened to Le Chat?

Le Chat is now called Vibe. Mistral's help center says the rename kept every account, plan and conversation, and chat.mistral.ai still works as the login. Vibe now bundles three modes: Vibe Work for multi-step tasks across your apps, Vibe Code for the CLI, VS Code and web, and Vibe Chat for the classic conversation view.

That matters for pricing because Pro is no longer a chat-only plan. The coding agent sits inside the same $14.99 subscription, so a developer who would otherwise pay for a separate coding assistant should compare Pro against that tool, not only against other chatbots. If you are weighing coding agents, our Claude Code pricing breakdown shows the other side of that comparison.

How much does Mistral Vibe cost?

The Vibe plans differ mostly in limits, not in which models you can use:

Read on Mistral's pricing page, 30 October 2026.

Pro also raises web searches to up to 5x Free and image generations to up to 40x. Team adds a little more headroom on top: Mistral's own example is 200 Flash answers a day on Team against 150 on Pro. Both paid plans can keep going past the plan limit with pay-as-you-go credits charged at API rates, which is a safety valve but also a way for a heavy month to cost more than the sticker.

The Team plan has a quiet floor. The seat counter on the pricing page starts at two users, so the smallest Team bill is $49.98 a month (our arithmetic: 2 × $24.99; the page rounds it to $50). A five-person team pays $124.95 a month on monthly billing, again our arithmetic. A solo user who only wants more storage should stay on Pro.

Is there an annual discount?

Yes. Mistral's subscription help article says Pro and Team can be billed monthly or annually, and annual billing is 20% cheaper than monthly. The pricing page shows only the monthly figures, so the exact annual charge appears at checkout. Cancel within 14 days of the first purchase and the plan ends at once, with a possible refund; after that, you keep access until the period ends.

How does the student price work?

The Education plan is Pro at $5.99 a month for verified students at accredited institutions, valid for up to 12 months and only if you have not used Vibe or Le Chat before. Mistral's student help article extends it to teachers and requires an institutional email address: student IDs and enrollment letters are not accepted. The Education card lists $15 a month in API credits instead of Pro's $25.5.

Does Mistral Pro include API credits?

Yes, on paper: the Pro card lists $25.5 a month in API credits, which is more than the plan's price. The catch is what those credits cover. Mistral's rate-limit help article separates two kinds of key: a plan key created from Vibe Code CLI draws on your plan's monthly Vibe budget and can stop working when that budget runs out, while a pay-as-you-go key created in the console bills per token under normal API rate limits.

So treat the Pro credits as fuel for Vibe and its coding tools, not as a free API budget for a product you ship. For an app in production, use a pay-as-you-go key and budget from the per-token rates below.

How much does the Mistral API cost per token?

The API is pay as you go, billed separately for input and output tokens. If tokens are new to you, what a token is in AI covers the unit. Here are Mistral's own generalist models on the standard tier:

Rates from Mistral's API pricing docs and model cards for Large 4 and Medium 3.5, read on 30 October 2026.

The naming misleads anyone who assumes size tracks price. Medium 3.5 is the most expensive model on the list, not a middle option: its output costs more than three times what Large 4 charges. Mistral recommends Medium for most tasks and coding, and it is generally available, while Large 4 launched on 6 October 2026 as a public preview with a 1M-token context window. Teams that need stability today pay the Medium premium; teams willing to test a preview get a much cheaper frontier model.

The pricing page FAQ still uses $0.5 in and $1.5 out as its "Mistral Large" example. Those are Large 3's rates, so check which Large you are calling before you budget.

Dupple data: of the eight Mistral generalist models in the table above, four charge $0.60 or less per 1M output tokens, and only Medium 3.5 charges more than $2.10, our count from Mistral's API pricing docs on 30 October 2026.

The same price list also carries third-party open models served by Mistral, such as DeepSeek V4.1 Flash at $0.30 input and $1.20 output and Z.ai GLM 5.3 at $1.40 and $4.40.

What does 1M or 10M tokens cost on Mistral?

Here is the monthly API bill at two volumes, assuming three input tokens for every output token, a common shape for chat and retrieval apps:

Our arithmetic: 75% of tokens at the input price, 25% at the output price, standard tier, no caching, no batch discount.

The spread is 30-fold between the cheapest and the priciest model at the same volume. For most internal tools and prototypes, even 10M tokens a month on Large 4 costs less than one Pro seat. The bill only gets serious on Medium 3.5 at volume, or when the output share rises: an app that writes long answers can flip the 3:1 split, and on Medium 3.5 output tokens cost five times input.

How do batch and caching cut the bill?

Two levers cut the bill on most models, and the first is the bigger one: batch processing costs 50% less, per the pricing page and the batch docs: you upload up to 1 million requests in one job and get results asynchronously, processed over a 24-hour window. On the Medium 3.5 example above, that takes a 100M-token month from $300 to $150 (our arithmetic). It suits classification, evaluations and overnight summaries, not anything a user waits on.

Cached input is the second lever, and it rewards apps that reuse the same instructions. Mistral says repeated prompt prefixes cost up to 90% less, and the docs list cached input at one tenth of the normal input rate on every generalist model. A long system prompt sent with every request is the classic case. Our guide on reducing LLM API costs lists the other levers, such as routing simple requests to a smaller model.

What do OCR, speech and voice agents cost?

Not everything bills per token:

From Mistral's API pricing docs and the OCR 4.1 model card, 30 October 2026.

At those rates, transcribing 100 hours of calls on Voxtral Mini Transcribe 3 costs $21, and parsing 10,000 pages with OCR 4.1 costs $40 (our arithmetic). Document and audio pipelines are cheap next to text generation; the summarization step that follows is usually the bigger line.

Is the Mistral API free?

Partly. Mistral's help center says every organization starts in Free mode, with the lowest rate limits, meant for evaluation and prototyping. Turning on pay-as-you-go unlocks Tier 1, and limits rise automatically as your cumulative bill passes $20, $100 and $500. Past $2,000 billed you contact support for higher limits. Buying prepaid credits does not raise your tier; only billed usage counts.

Priority Tier, with a 99.5% uptime SLA and its own queue during peak load, needs an agreement with Mistral and has no public price. And because several Mistral models ship as open weights, self-hosting is a third route: no per-token bill, but you pay for GPUs and should check each model's license, since Mistral says some commercial deployments need a separate license. Our list of free local LLM tools covers running small models on your own machine.

Is Mistral worth it compared with other options?

For the app, Pro makes sense if you want one subscription that covers chat, documents, slides and a coding agent, and you like that Mistral is a European provider. Its weakness is limits stated only as multiples of the free plan, so you cannot plan heavy use from the pricing page alone. Our ChatGPT alternatives roundup puts it next to the other assistants, and AI Academy's Le Chat review goes deeper on the app itself.

For the API, Large 4 and Small 4 are priced aggressively for their size; Medium 3.5 is the one to cost carefully. If you want to switch between Mistral and other providers per request, an LLM gateway or OpenRouter does that, and Together AI's pricing shows what a GPU-hour host charges for open models. Builders wiring Mistral into retrieval or agents can start from the best RAG tools and the best AI agent frameworks.

How we checked these prices

We read Mistral's pricing page, the API pricing docs, the model cards linked above, the Priority Tier docs and four help center articles on 30 October 2026, with prices shown in US dollars. Scenario volumes, the 3:1 input-to-output split and every total marked as ours are our own arithmetic from list prices. Mistral renames products and adds models often, so check the model card before you budget. A summary of the plans also sits on Mistral AI's Toolradar pricing profile.

FAQ

Bottom line

Buy Vibe Pro if you want Mistral's assistant and coding agent in one $14.99 subscription; move to Team only when you have at least two people who need a shared workspace. For the API, start on Small 4 or Large 4, keep Medium 3.5 for the tasks that need it, and use batch for anything that can wait.

Cite this: Dupple, "Mistral AI Pricing 2026: Vibe (Le Chat) Pro, Team and API Costs per Token", November 2026.

Pricing out the AI stack for your team? Start the Dupple X trial.