Mistral is Europe's frontier AI lab, built on open weights, EU data residency and unusual efficiency for its size. Its consumer assistant Le Chat — rebranded as Vibe — sits at $14.99 a month for Pro against $20 from the US labs, and the API is aggressively priced with Mistral Large 3 at roughly $0.50 in and $1.50 out per million tokens. The reason to take it seriously isn't benchmark leadership, which it doesn't claim: it's that ASML took an 11% stake, ARR reached around $400 million, and for European organisations the jurisdiction question has an answer that isn't American.
Reviewed by EdGrows·Updated Aug 18, 2026Affiliate link — we may earn a commission
This analysis is based on documentation, public user reports, and vendor materials — not yet on our own hands-on testing. How we rate
Mistral doesn't claim to have the best model, and pretending otherwise would be the fastest way to misjudge it.
The argument is structural. If you are a European bank, hospital, government department or defence supplier, "send our data to an American company under American jurisdiction" is frequently not an option your legal team will sign. Until recently, the alternative was no AI at all.
Mistral is the alternative. Paris-based, founded in 2023 by three researchers from Google DeepMind and Meta, built on open weights, EU data residency and on-premise deployment, with models good enough for the overwhelming majority of real work.
That's the whole pitch, and the market has responded: reported ARR of roughly $400 million in early 2026, up from around $16 million at the end of 2024.
Mistral's September 2025 Series C raised €1.7 billion, led by ASML, which put in €1.3 billion for roughly 11% and became its largest shareholder, valuing the company at about €11.7 billion.
ASML is not a generalist investor. It is the world's only manufacturer of advanced EUV lithography equipment — a genuine chokepoint in global chip production. Its investment came with a strategic partnership, which ties Mistral to Europe's technology-industrial base rather than to a fund looking for an exit.
For an enterprise buyer, that's a more useful signal than any benchmark score. The question "will this vendor exist in five years" gets a better answer from a shareholder like that than from a model leaderboard that changes monthly.
Further raises at higher valuations were reported during 2026, including talk of rounds well above the Series C mark, though not all were confirmed closed. The direction is clear enough; the specific numbers circulating should be treated with care.
Le Chat launched in February 2024 and became a genuine national moment in France — President Macron publicly urged citizens to download it instead of ChatGPT, which is not a marketing beat most AI assistants get.
It has since been rebranded as Vibe. Same product, new name, and worth knowing when you encounter older documentation.
Pricing undercuts the field: free tier, Pro at $14.99, Team at $24.99 per user. Against $20 for ChatGPT Plus, Claude Pro and Google AI Pro, that's a 25% discount on the going rate for a mainstream assistant.
Whether the discount matters depends on what you'd be giving up. For everyday drafting, summarising, questions and light coding, very little. For frontier reasoning, long agentic chains and the hardest technical work, the US labs remain ahead — and if that's your daily workload, the $5 saving is not the deciding factor.
| Model | Input / output per 1M tokens |
|---|---|
| Mistral Small 4 | ~$0.15 / $0.60 |
| Mistral Large 3 | ~$0.50 / $1.50 |
| Mistral Medium 3.5 | ~$1.50 / $7.50 |
Large 3 at $0.50 in and $1.50 out is aggressive pricing for a model of its class, and the efficiency story behind it is real — Mistral has consistently done more per parameter than labs with far larger compute budgets.
The oddity in that table is worth flagging: Medium 3.5 costs considerably more than Large 3. The naming does not map to price the way you'd assume, so check the current rate card against the model you actually intend to call rather than assuming bigger means dearer.
Add the open-weight option and you get something no US frontier lab matches: run it yourself on your own hardware, or buy an enterprise contract with private and on-premise deployment and support. For regulated organisations, that combination is the product.
You want the strongest model, full stop. Claude, ChatGPT and Gemini hold the frontier. Mistral is competitive, not leading.
You need a deep ecosystem. Plugins, integrations, community tooling — the US labs have years of accumulated third-party work that Mistral doesn't.
Jurisdiction is irrelevant to you. Strip out the sovereignty argument and you're choosing a slightly cheaper, slightly weaker assistant. That's a reasonable trade, not a compelling one.
You want free. DeepSeek costs nothing at all, though it answers the jurisdiction question in the opposite direction.
Against ChatGPT: ChatGPT is stronger, has a far wider ecosystem, and costs $19.99. Mistral is $14.99, open-weight, EU-resident and self-hostable. Where data residency is a requirement this comparison ends immediately; where it isn't, ChatGPT wins on capability and tooling.
Against Claude: Claude is the better writing and reasoning partner and offers enterprise deployment of its own. Mistral's differentiator remains jurisdiction and the ability to run models on your own infrastructure.
Against DeepSeek: DeepSeek is the other major open-weight lab, free for consumers and cheaper on API. It's also Chinese, which means for European organisations it fails the same test that made them look at Mistral in the first place. Two open-weight answers to two incompatible jurisdiction questions.
Against Gemini: Gemini bundles Workspace integration, video generation and storage at $19.99. Mistral bundles none of that and offers something Google structurally cannot: a model you can run inside your own perimeter under EU law.
Against Microsoft Copilot: Microsoft Copilot is the incumbent inside European enterprises through Office. Mistral increasingly appears alongside it in sovereignty-sensitive departments rather than replacing it wholesale.
| Product | Price | Notes |
|---|---|---|
| Le Chat / Vibe Free | $0 | Consumer assistant, capped usage |
| Le Chat / Vibe Pro | $14.99/mo | Undercuts the $20 US standard |
| Le Chat / Vibe Team | $24.99/user/mo | Collaboration features |
| Enterprise | Custom | Private, on-premise and sovereign deployment |
| API — Small 4 | ~$0.15 / $0.60 per 1M | Cheapest tier |
| API — Large 3 | ~$0.50 / $1.50 per 1M | Flagship, aggressively priced |
| API — Medium 3.5 | ~$1.50 / $7.50 per 1M | Costs more than Large 3 despite the name |
| Open weights | Free | Self-host, permissive licences on much of the family |
Checked August 2026 against Mistral's published pricing and current reporting. The consumer assistant Le Chat has been rebranded as Vibe. Valuation figures cited are from the confirmed September 2025 Series C (€1.7B led by ASML, ~€11.7B post-money); larger rounds reported during 2026 were not all confirmed closed. Model names do not track price order — verify the current rate card at mistral.ai/pricing before budgeting.
Decide whether jurisdiction is a requirement or a preference. If it's a requirement, Mistral is on the shortlist by default and the comparison is against other EU options. If it's a preference, judge it as a cheaper assistant and weigh the capability gap honestly.
Check the API rate card per model, not by name. Medium 3.5 costing more than Large 3 is the kind of detail that quietly inflates a bill.
Take the open weights seriously if you have hardware. Self-hosting removes the residency question entirely and gives fixed costs at volume — the same argument that makes DeepSeek interesting, with a European answer.
Pro at $14.99 is a fair consumer price. Not the best assistant available, and 25% below the going rate for something that handles most work fine.
Mistral is the only European AI lab operating at genuine frontier-adjacent scale, and in 2026 it stopped being a hopeful national champion and became a real business — roughly $400 million in ARR, ASML holding 11% as its largest shareholder, and enterprise deployments across European governments and industry. For organisations where data residency is a legal requirement rather than a preference, it's frequently the only credible option, and that's a strong position regardless of benchmarks.
The honest limits: it doesn't lead on capability. Claude, ChatGPT and Gemini are ahead at the frontier, the ecosystem around Mistral is thinner, and if you strip out the sovereignty argument you're left with a slightly cheaper, slightly weaker assistant. The naming has shifted too — Le Chat is now Vibe — and the API rate card doesn't follow the intuition that bigger models cost more.
What makes it worth recommending anyway is the combination nobody else offers in Europe: open weights you can self-host, a vendor you can sign an on-premise contract with, and a company under EU jurisdiction with an industrial shareholder base. That is a different product from a chatbot subscription, and it's the reason regulated European buyers keep choosing it.
For European organisations with residency requirements, recommend and evaluate seriously. For individuals choosing an assistant on capability alone, the US labs remain ahead — though at $14.99 the gap is narrower than the price difference suggests.
Note: Mistral does not currently have an affiliate program with AIVario. We earn no commission, and this rating carries no commercial incentive.
Best for: European organisations with data-residency requirements, regulated industries and public sector, developers wanting open weights with commercial support, self-hosted deployments Not ideal for: Users who want the strongest available model, ecosystem-dependent workflows, anyone for whom jurisdiction is irrelevant, users wanting a free tier Bottom line: Europe's credible frontier lab, cheaper than the US assistants and self-hostable under EU law — bought for sovereignty and efficiency rather than for topping leaderboards.
Le Chat has a free tier, Pro at $14.99 a month and Team at $24.99 per user per month, with custom enterprise pricing for private and on-premise deployments. That undercuts ChatGPT Plus, Claude Pro and Google AI Pro, all at around $20. API pricing is separate and per token: Mistral Large 3 at roughly $0.50 input and $1.50 output per million tokens, Medium 3.5 at $1.50 and $7.50, and Small 4 at $0.15 and $0.60.
Yes — the consumer assistant has been rebranded as Vibe. The underlying product is the same assistant that launched in February 2024 and became a national talking point in France, to the point that President Macron publicly urged citizens to use it instead of ChatGPT. If you find documentation or comparisons referring to Le Chat, they're describing the same thing under its previous name.
Close enough for most work, not ahead. Mistral Large 3 and the Magistral reasoning models handle mainstream tasks well, and the efficiency-per-parameter is genuinely impressive for a lab with a fraction of the compute budget. Where it trails is at the frontier — the hardest reasoning, the longest agentic chains, the most demanding code. The honest framing is that Mistral is good enough for the overwhelming majority of real tasks and rarely the best available model for any single one.
Data residency inside the EU, a company under EU jurisdiction and law, and deployment options that keep everything on infrastructure you control. For a European bank, hospital, government department or defence supplier, that isn't a marketing preference — it's frequently the difference between being able to deploy AI at all and not. For everyone else it's a nice-to-have that costs nothing extra, since Mistral is also cheaper than the US alternatives.
Many are open-weight, which is not quite the same thing but is the meaningful part. Mistral publishes weights under permissive licences for a substantial portion of its family, letting you download, run and fine-tune them on your own hardware. Its premium models stay closed. The strategy is deliberate: open weights drive developer adoption, which funnels serious usage into the paid API and enterprise contracts. It's the same playbook that made the lab's early reputation.
Europe's largest AI company by a distance, and small next to the US labs. Its September 2025 Series C was €1.7 billion led by ASML, which invested €1.3 billion for roughly 11% and became its largest shareholder, valuing the company at about €11.7 billion. Reported ARR reached roughly $400 million in early 2026, up from around $16 million at the end of 2024. Further raises at higher valuations were reported during 2026 — treat unconfirmed figures with care.
Yes, and this is a genuine differentiator. Open-weight models can be self-hosted outright, and Mistral sells enterprise contracts covering private and on-premise deployment with support. That combination — a frontier-adjacent model you can run inside your own perimeter, from a vendor you can sign a contract with — is something neither OpenAI nor Anthropic offers in the same form, and it's the core of Mistral's enterprise business.
Because of who ASML is. The Dutch company is the world's only manufacturer of advanced EUV lithography equipment — effectively a chokepoint in global semiconductor manufacturing. Its €1.3 billion investment and strategic partnership tie Mistral into the heart of Europe's technology-industrial base rather than to a financial investor. For enterprise buyers evaluating whether Mistral will still exist in five years, that shareholder register is a more useful signal than any benchmark.
The leading GTM data platform, repriced from the ground up in March. Two plans now, two credit types, and a lot of stale advice floating around.
The category leader in revenue intelligence, priced accordingly. The per-seat number is the smallest part of what you'll actually pay.
The best backlink data in the industry, sold with a credit meter and single-seat plans that make the real cost hard to read.