Mistral suits organisations that want open-weight models and European infrastructure they can control. Its €3 billion Series D in September 2026 funds training in its own datacenters. The catch: Large 4 weights and licence were still pending in early October 2026.
Mistral AI is a Paris lab founded in April 2023 whose newest model, Mistral Large 4, has 1 trillion parameters with 52 billion active per token. It sells models, the Vibe app, Studio, Inference, Compute and Forge, and positions itself as a European alternative to US-hosted AI providers.
Founded: 2023 · HQ: Paris, France · Team: 150+ · CEO: Arthur Mensch · Funding: €3B Series D (Sept 2026); €1.7B Series C (Sept 2025); $830M debt (Mar 2026) · Valuation: Over €21B post-money (Series D, Sept 2026)
About Mistral AI
Mistral AI was founded in April 2023 in Paris by Arthur Mensch (CEO), Guillaume Lample (Chief Science Officer) and Timothée Lacroix (CTO). Mistral's own timeline dates the first hire to 5 June 2023 and the 100th employee to 30 September 2024. The company opened a German hub in Munich on 28 September 2026 and says it now operates across 20 countries with 125+ enterprise customers, including Airbus, ASML and HSBC.
The first release, Mistral 7B on 27 September 2023, was open-weight and carried a permissive Apache 2.0 licence, which built the developer following that the paid API later drew on. Mixtral 8x7B followed in December 2023 and showed the mixture-of-experts idea at a usable size: eight experts, two active per token. Mistral Large arrived in February 2024 as the first closed flagship, and the line has since grown into smaller edge models such as Ministral 3 8B, the document model Mistral OCR 4 and the previous flagship Mistral Large 3.
The current flagship is Mistral Large 4, announced as a public preview on 6 October 2026. Mistral describes it as a natively multimodal model with 1 trillion parameters and 52 billion active, trained on 3,800 NVIDIA Grace Blackwell GPUs in its own European datacenters. The weights are promised for the end of October 2026, and the licence had not been announced at the time of this review. Our own explainer, what Mistral Large 4 is for, covers where it fits.
Funding explains the strategy. Mistral's September 2025 Series C raised €1.7 billion at an €11.7 billion post-money valuation, led by ASML. On 8 September 2026 it announced a €3 billion Series D at a post-money valuation above €21 billion, led by Samsung Electronics with the Scaleup Europe Fund (managed by EQT) and PSG Equity as co-leads. In March 2026 it also closed its first debt financing, $830 million from a seven-bank group, to buy 13,800 NVIDIA GB300 chips for a datacenter near Paris, as reported by Reuters and others.
Today the company sells a stack rather than only a model. Its products page lists Vibe (agents for research and everyday work, covered on the Mistral AI tool page), Studio for building and governing agents, Inference for open-weight and third-party models on one API, Compute for reserved GPUs in Europe, and Forge for customising and training models on an organisation's own data. Recent partner announcements include HUMAIN (24 August 2026), Cloudera (10 September 2026) and Mozilla's Firefox Smart Window (16 September 2026).
Where Mistral is the weaker choice: buyers who want the longest-established enterprise search stack or a single closed model with the most third-party tooling often look at Cohere, OpenAI or Anthropic first, and teams that mainly want the lowest open-weight price point compare DeepSeek. AI21 Labs is a narrower alternative for long-context text work. The open-weight promise also depends on release timing: until the Large 4 weights actually ship, the preview is an API-only product, and Mistral says a separate group of vetted partners gets the model with reduced moderation for security work.
For a broader view of how Mistral sits among the other labs, see our best AI companies guide. Facts on this page were read from Mistral's own news, about and products pages on 10 October 2026; the fundraising history and debt financing are cross-checked against press coverage.
Mission
To develop open-source and commercial language models emphasizing efficiency, privacy, and responsible deployment while maintaining values of transparency and accessibility.
Products
- Mistral Large 4
- Vibe
- Studio
- Inference
- Compute
- Forge
Mistral AI Models on HokAI
Compliance
SOC 2 Type II
Links
Frequently Asked Questions
What does Mistral AI sell in 2026?
Mistral sells a full stack: models such as Mistral Large 4 and Ministral 3, the Vibe agent app, Studio for building agents, Inference for API access, Compute for reserved European GPUs and Forge for custom model training. Open-weight releases remain part of the pitch, with Large 4 weights promised for the end of October 2026.
How much has Mistral AI raised and what is it worth?
The latest round is a €3 billion Series D from 8 September 2026, valuing the company at more than €21 billion once the money is counted. The previous round, in September 2025, was a €1.7 billion Series C at €11.7 billion. Older figures such as $6 billion come from the 2024 Series B.
Is Mistral Large 4 open-weight yet?
Not at the time of this review. Mistral launched a public preview API on 6 October 2026 and says weights follow by the end of that month, but it had not published the licence. Check the model page before building a self-hosted plan on it.
How does Mistral AI compare with Cohere in 2026?
Mistral competes on open weights and European infrastructure it runs itself, while Cohere focuses on closed models for enterprise search and private deployment. Pick Mistral when you want weights you can host and modify; pick Cohere when a packaged retrieval stack matters more.
How do you start using Mistral AI?
Sign in to Studio at console.mistral.ai to get an API key and try the Large 4 preview, or open Vibe at chat.mistral.ai for the app. Smaller open-weight models can be downloaded from the mistralai organisation on GitHub and Hugging Face for self-hosting.
Top Alternatives
HokAI guides covering Mistral AI
- What Mistral Large 4 Is Actually For: Mistral Large 4 is a 1.05T open-weight preview with a halved sale price. Who it suits, what it costs, and what is still unknown about weights and licence.
- Which LLM Should You Use? GPT-6.1 Sol, Opus 5.5, Argon and More (October 2026): Claude Opus 5.5, GPT-6.1 Sol, Haiku 5.5 and Mistral Large 4 compared on price and independent scores, with which you can buy today. Checked 9 October 2026.
- The Cheapest LLM APIs in 2026, and What Each Cheap Price Leaves Out: Compare six cheap LLM APIs on price per million tokens, quality score and fine print: Claude Haiku 5.5, GLM-5.3 Flash, Ling, Gemini, DeepSeek and Mistral.
- Grok 4.7 vs Claude Opus 5 vs GPT-6 Astra vs Gemini 3.8 Flash: What the Launch Numbers Don't Tell You: Grok 4.7, Claude Opus 5, GPT-6 Astra and Gemini 3.8 Flash all launched within seven weeks. Here is what their price, speed and benchmark numbers actually mean.
- Best AI Companies in 2026: Who Is Actually Leading: Anthropic's revenue passed OpenAI's in 2026, even as OpenAI keeps the bigger valuation. See how six leading AI companies actually compare, by the numbers.
- Nvidia Just Bought Hugging Face for $12.93 Billion. Here's What Actually Changes.: Nvidia is buying Hugging Face for $12.93 billion, closing in 2027. Here is the concrete tripwire to watch, and the smaller precedent nobody is citing yet.