Mistral AI Pricing 2026: What Vibe (Le Chat) and the API Actually Cost

Mistral AI pricing runs on two separate tracks, and the headline number is hard to beat. Vibe, the consumer app Mistral used to call Le Chat, starts at $0/month and runs to $14.99/month for Pro and $24.99/user/month for Team, while the developer API charges per million tokens, starting at $0.10. That makes Mistral one of the cheapest ways to reach a frontier model in 2026, and the open-weight Mistral Large 3 undercuts GPT-6.1 Sol by 75% on input and 85% on output.

This guide breaks down every price you care about, the four Vibe plans, the full API rate card per million tokens, what the free tier actually gives you, and how the numbers stack up against ChatGPT, Claude, and DeepSeek. We re-verified every figure against Mistral’s plans page and its separate API rate card on 25 September 2026, because most third-party trackers are months out of date and still quote retired rates and a product name Mistral no longer uses. If you want Mistral plus several other models without juggling subscriptions, we cover that route too. And if you saw the viral Le Chaton Fat model making the rounds, that one is a hoax; everything below is Mistral’s real, current pricing.

The Key Takeaways

• Vibe Free costs $0/month with real frontier-model access plus $10/month in API credits; Mistral no longer publishes a message number, only "limited messages and web searches"
• Vibe Pro is $14.99/month, one of the cheapest premium AI chat subscriptions from a major lab, and Le Chat is its old name
• The Mistral API starts at $0.10 per million input tokens and runs to $1.50 in / $7.50 out on Mistral Medium 3.5
• Mistral Large 3 costs just $0.50 input / $1.50 output per million tokens, 75% and 85% under GPT-6.1 Sol, but GPT-6 Luna at $0.10 / $0.50 now undercuts it on both
• Verified students get Pro for $5.99/month, and every Vibe plan now bundles monthly API credits: $10 Free, $30 Pro, $50 Team

How Mistral AI Pricing Works: Two Separate Tracks

The single biggest source of confusion is that Mistral sells two different things that are billed independently. The first is Vibe, the consumer app you use in a browser, on mobile, in your terminal or in your IDE, sold as a flat monthly subscription. Mistral renamed it from Le Chat in August 2026; the URL is still chat.mistral.ai, and most price trackers have not caught up. The second is the API on La Plateforme, which developers pay for by the token based on how much text goes in and out.

This used to be a hard wall, and it is not one any more. Every Vibe plan now includes a monthly allowance of API credits, $10 on Free, $30 on Pro and $50 on Team, so a subscription does buy you a small amount of API usage. Past that allowance the two are still billed separately: topping up your API balance does nothing for your chat limits, and the credits do not roll over as a discount on the subscription. If you just want to chat, generate images, and run code, you want a Vibe plan. If you are building an app, a chatbot, or an automation at any real volume, you want the API.

Coding is not a separate purchase either. Vibe for code, which runs in the terminal as Vibe CLI and as a VS Code and JetBrains plugin, is bundled into the same plans as the chat interface, so the Pro subscription covers chat, image generation, deep research, and all-day coding under one price.

Vibe Pricing Plans (Formerly Le Chat)

Vibe comes in four tiers, from a genuinely useful free plan up to custom enterprise deals. Here is what each one costs and unlocks, with prices read off Mistral's plans page on 25 September 2026.

Vibe Free

The free tier is more generous than most rivals. You get real access to Mistral’s current models, image generation, coding sessions in the CLI, more than 100 connectors and $10/month in API credits for testing in Studio. Mistral does not publish a number for the free message limit, only "limited messages and web searches", and it prices Pro against it as "up to 6x free" on messages and "up to 5x free" on web searches. Treat any specific daily figure you see quoted elsewhere, including the 25-a-day number this guide used to carry, as unsourced.

Vibe Pro

Pro at $14.99/month is the sweet spot, and it is one of the cheapest premium chat plans any major lab offers. Mistral sizes it against the free plan directly: up to 6x the messages, up to 5x the web searches, up to 40x the image generations, a 15GB library, unlimited task scheduling, and all-day coding in the CLI, the IDE or on the web. It also carries $30/month in API credits, double what Free gets. Pro can opt out of model training, which stops Mistral using your prompts and outputs.

Vibe Team and Enterprise

Team at $24.99/user/month adds a shared workspace, 30GB of storage per user, domain-name verification, data export and $50/month in API credits, which suits small startups that want one billing relationship. Enterprise pricing is custom and aimed at companies that need private deployment, custom models, SAML single sign-on, audit logs, and dedicated support. You contact Mistral sales for a quote.

Student Discount

Verified students can get Vibe Pro for $5.99/month instead of $14.99, valid for up to 12 months, and only if you have not used Vibe or Le Chat before. The education plan carries $15/month in API credits rather than Pro's $30. If you are at university, that is a hard discount to argue with, and it pairs well with other AI free trials and student deals worth stacking.

Mistral API Pricing Per Million Tokens

The Mistral API charges by the token, split between input (your prompt) and output (the model’s reply), priced per million tokens. The lineup runs from tiny edge models at a few cents up to the multimodal flagship. Batch processing earns a 50% discount on most models.

Three rows changed in September 2026 and are worth calling out. Devstral 2 and the first-generation Ministral 8B and 3B are gone from the rate card: Devstral 2 was retired on 31 July 2026 with Mistral Medium 3.5 named as its replacement, and the Ministral line was replaced by Ministral 3 at the same prices. Mistral now also resells Z.ai’s GLM 5.3 through its own API. Mistral does not print context windows on the rate card; Artificial Analysis measures 256K for Mistral Large 3, Mistral Small 4 and the whole Ministral 3 line, and Z.ai gives GLM 5.3 a 1M-token window.

Notice that Mistral Large 3 is three times cheaper than Mistral Medium 3.5 on paper, which is the opposite of what the names suggest. Mistral is not entirely consistent about which of the two is its flagship: the API rate card calls Large 3 the "flagship multimodal and multilingual model", while the models overview calls Medium 3.5 "our frontier-class multimodal model" and the pricing FAQ answers its own "which model should I use?" with "for most tasks and coding: Mistral Medium". The tiebreaker is the release dates. Large 3 is version 25.12, Medium 3.5 is 26.04, so Medium is the newer model and the one Mistral points builders at. Take Large 3 when price per token matters more than quality, and Small 4 for high-volume, simple jobs. Since October 6, 2026 the top of the range has a clearer answer: Mistral Large 4, now in preview at $1.36 input and $4.18 output, is what Mistral calls its "largest and most capable model to date".

Reasoning and Specialized Models

Beyond the core lineup, Mistral prices a set of specialists separately, and the list has thinned. The Magistral reasoning family and Pixtral are both off the rate card entirely, retired into Mistral Medium 3.5; if a comparison page still quotes a Magistral rate, it is quoting a dead model. What is live: OCR 4.1 at $4 per 1,000 pages for plain extraction and $5 per 1,000 for Document AI, Voxtral Mini Transcribe 2 at $0.003 per minute of audio and Voxtral TTS at $0.016 per 1,000 characters, and the embedding models Mistral Embed at $0.10 and Codestral Embed at $0.15 per million input tokens. Mistral Moderation 2 and the Lean 4 agent Leanstral are free. Agent tools bill per call rather than per token: web search and code execution are $30 per 1,000 calls, image generation $100 per 1,000 images, and premium news $50 per 1,000 calls. If you only need chat or simple generation, you can ignore all of these.

Is There a Free API Tier?

Not a standing one, but the free plan now carries $10/month in API credits, which is the practical equivalent for prototyping. Pro raises that to $30 and Team to $50. Past the allowance you are on paid per-token rates, and Mistral offers a 50% batch discount, 90% off cached input tokens on repeated prompts, a 10% surcharge for EU regional processing, and an Enterprise tier at 75% above list. For day-to-day chatting rather than building, the Vibe Free plan is still the better no-cost option.

How Mistral Pricing Compares to ChatGPT, Claude, and DeepSeek

The reason Mistral keeps showing up in cost-cutting conversations is simple. Its flagship is priced like a budget model while performing near the top tier. Here is Mistral Large 3 against the comparable flagships from OpenAI and Anthropic.

The takeaway has changed, and it is worth being blunt about it. Against the flagships, Mistral Large 3 is still a bargain: it is 20x under GPT-6 Astra on input and 33x on output, and 8x and 13x under Claude Opus 5.5. What is no longer true is that Mistral owns the floor. GPT-6 Luna at $0.10 / $0.50 undercuts Large 3 on both axes, five times cheaper on input and three times on output, which reverses the "nothing from OpenAI comes close" line this guide used to run. DeepSeek sits between them: it switched to peak and off-peak billing on 16 August 2026 and cut the Flash rate again on 10 September, so V4.1 Flash costs $0.15 / $0.60 off-peak and double that during peak hours (01:00-04:00 and 06:00-10:00 UTC, Monday to Friday). Off-peak it beats Large 3 and loses to Luna. Mistral still belongs in our Qwen pricing and MiniMax pricing breakdowns of cheap-but-capable models, but the bottom of this table is now a contest it does not win. One correction worth making explicitly: GPT-6 Astra is callable through the API, as gpt-6-astra, and has been since shortly after its 3 September 2026 launch. For a wider field, our DeepSeek pricing guide and Anthropic vs OpenAI cost comparison go deeper on the trade-offs.

Want Mistral and Every Other Model in One App?

The catch with single-lab pricing is that no one model wins every task. Mistral Large 3 is brilliant value, but you might want Claude for long writing, ChatGPT for general questions, Gemini for research, or DeepSeek when cost matters most. Paying for several subscriptions adds up fast.

This is where a multi-model app changes the maths. Fello AI gives you Mistral, Claude, ChatGPT, Gemini, Grok, and DeepSeek through one $9.99/month subscription on your Mac, so you can run the same prompt across models and keep the best answer. Mistral has no Mac app of its own, so the Mistral desktop client in Fello AI is also a native way to use it on a Mac. You can get started with Fello AI and stop guessing which lab to commit to. For a model-by-model rundown of what each one is good at, see our best AI models guide.

Conclusion

Mistral AI pricing is still among the most aggressive in the market, with one caveat it did not have three months ago. Vibe, the app formerly called Le Chat, is free for casual use or $14.99/month for Pro, and the API starts at $0.10 per million tokens with open-weight models that undercut GPT-6.1 Sol and Claude by wide margins. The caveat is the floor: OpenAI's own GPT-6 Luna is now cheaper than anything Mistral sells. If you want a cheap, private, capable AI, care about open weights and EU hosting, and are comfortable outside the OpenAI and Anthropic ecosystems, Mistral is one of the best-value picks of 2026. If you would rather not bet on a single lab, a multi-model app like Fello AI gets you Mistral-class value plus four other top models for less than the price of one premium subscription.

FAQ