Mistral AI
Paris model lab with open weights, cheap APIs and the Vibe agent
Overview
Mistral AI is a Paris-based AI company that builds open-weight and proprietary language, vision, speech and OCR models, sells them through a pay-per-token API, and runs Vibe, its assistant for work and coding (formerly Le Chat). Its October 2026 headline model is the Mistral Large 4 preview.
Pricing
Pricing shown for reference only. These figures reflect RECATOOLS research as of 14 Oct 2026 and may be out of date or incomplete. This is not financial or purchasing advice — always confirm the current price on the provider’s official website before making any decision.
- Limited messages and web searches
- Limited coding sessions
- Basic image generation and editing
- 100+ connectors
- More messages and web searches
- All-day coding in CLI, IDE or web
- Spreadsheets and slides
- Monthly API credits
- Chat and email support
- Even higher limits
- Up to 30GB storage per user
- Domain name verification
- Data export
- Custom models, agents and workflows
- Audit logs
- SAML SSO
- White label
- Batch processing at 50% off
- Cached input up to 90% off
- OCR priced per 1,000 pages
API pricing (pay as you go)
| Model | Input | Cached input | Output |
|---|---|---|---|
| Mistral Medium 3.5 | $1.50 | — | $7.50 |
| Mistral Large 3 | $0.50 | — | $1.50 |
| Mistral Small 4 | $0.15 | — | $0.60 |
| Ministral 3 (14B) | $0.20 | — | $0.20 |
| Ministral 3 (8B) | $0.15 | — | $0.15 |
| Ministral 3 (3B) | $0.10 | — | $0.10 |
| Codestral | $0.30 | — | $0.90 |
| GLM 5.2 Third-party model served on Mistral's API | $1.40 | $0.14 | $4.40 |
Use cases
What you can produce with Mistral AI
- Self-hosted deployments of Apache 2.0 models such as Small 4, Large 3 and Ministral 3
- Low-cost API inference, from $0.10 per million tokens on Ministral 3 3B
- Agentic coding sessions through the Vibe CLI, VS Code and JetBrains plugins and remote agents
- Long multi-step work tasks across inbox, calendar and connectors in Vibe Work Mode
- Document text and structure extracted with Mistral OCR
- Speech transcription and text-to-speech with the Voxtral models
Mistral is a good choice when you want models you can host yourself or buy cheaply by the token: Small 4, Large 3 and the Ministral 3 models are Apache 2.0, and API rates sit well below most US frontier labs. The Vibe agent now bundles chat, research, documents and a coding agent across CLI, IDE and web for $14.99 a month.
Read each model's licence rather than trusting the "open" label, since Medium 3.5 uses a modified MIT licence and Mistral's own FAQ adds commercial terms. Large 4 is still a preview until its weights ship. Worth shortlisting for cost-sensitive or self-hosted deployments.
What we found
Mistral AI is a Paris company that released its first model, Mistral 7B, in September 2023 and its chat assistant, Le Chat, in February 2024. On 28 May 2026 Le Chat became Vibe, one agent for both office work and coding, with Work and Code modes, a command-line tool, VS Code and JetBrains plugins, and every conversation and plan carried over. On 6 October 2026 Mistral opened a public API preview of Mistral Large 4, a natively multimodal model with one trillion parameters, 52 billion of them active per token, trained on Mistral's own European GPUs; the company says the weights follow by the end of October.
Licences need checking model by model. Mistral's model overview lists Small 4, Large 3 and the Ministral 3 models under Apache 2.0, Medium 3.5 under a modified MIT licence, Voxtral TTS under a non-commercial licence, and Codestral, OCR and the embedding models as proprietary. Its own pricing FAQ muddies this by saying commercial deployment of open-weight models requires a Mistral licence, which sits oddly with Apache 2.0, so read the card for the model you plan to ship. The Magistral reasoning and Devstral coding lines are now on the retired list, with Medium 3.5 and Small 4 named as replacements.
API prices are low. Large 3 costs $0.50 in and $1.50 out per million tokens, Medium 3.5 $1.50 and $7.50, Small 4 $0.15 and $0.60, and Ministral 3 3B $0.10 each way, with Large 4 at a preview sale price of $0.68 and $2.09. Batch jobs cost half, and cached input up to 90% less. Vibe is free with limits; Pro is $14.99 a month before tax, or $5.99 for verified students, and Team is $24.99 per user.
Mistral announced a €3 billion Series D on 8 September 2026 at a post-money valuation above €21 billion, led by Samsung Electronics, and says it serves more than 125 enterprises. For ASEAN teams the draw is self-hostable weights at low API rates; its sovereignty pitch is built for Europe.
Compiled by RECATOOLS from public sources: vendor documentation and pricing, code repositories and published reporting.
Notable facts
- Mistral 7B was released via a BitTorrent magnet link — an unusual distribution method that made it impossible for the company to take the model back once released.
- All three Mistral co-founders previously worked at Google DeepMind or Meta AI Research, making the team among the most experienced in European AI.
- Mixtral 8x7B is technically a 47-billion-parameter model, but only 13B parameters are active per inference, making it extremely cost-efficient to run.
Frequently asked questions
About this listing
This entry was compiled from publicly available data including Mistral AI's official website, press releases, documentation, and reputable third-party publications. RECATOOLS is not affiliated with Mistral AI unless explicitly stated.
Third-party AI tools update their pricing, features, availability, and policies frequently. Information here may be outdated by the time you read this — we make reasonable efforts to keep listings current, but cannot guarantee absolute accuracy.
For the latest details, please refer to Mistral AI directly →
Spotted something out of date? Suggest an update →
Alternatives to Mistral AI
More in LLMs & Chat
