Mistral AI

What Mistral AI does

Mistral AI is a Paris-headquartered artificial intelligence company founded in 2023 by former Google DeepMind and Meta researchers Arthur Mensch (CEO), Guillaume Lample (Chief Scientist), and Timothée Lacroix (CTO). The company builds, trains, and deploys a portfolio of proprietary foundation models spanning text, code, image, audio, and structured data, including Mistral Large 3, Mistral Medium 3.5, Mistral Small 4, Codestral, Devstral, Voxtral, Magistral, and the OCR 4 document intelligence model. It exposes these models through a full-stack platform consisting of Mistral Studio (production and orchestration), Mistral Forge (custom training and alignment), Mistral Compute (sovereign GPU cloud on NVIDIA Grace Blackwell systems), and the Vibe agent product (rebranded from Le Chat in May 2026) for chat, code, and long-horizon enterprise work. Open-weight releases (Mixtral 8x7B, 8x22B, Mistral Small 4 under Apache 2.0, and Devstral 2) are paired with proprietary frontier models and on-premises deployment options, all marketed under a European AI sovereignty positioning that is reinforced by Mistral Compute's EU-hosted infrastructure and explicit GDPR, DORA, and EU AI Act alignment.

Mistral serves a mix of large enterprise and public-sector customers, with named adopters including HSBC, BNP Paribas (65,000 users on BNP's own infrastructure), AXA (140,000+ employees), ASML, Airbus (full Mistral suite, on-premises), BMW, Stellantis, the French Ministry of Defense, France Travail, CMA CGM, Cisco, TotalEnergies, Mozilla (Firefox Shake-to-Summarize), and the Singapore HTX agency. Revenue is generated through a combination of usage-based API pricing (per million tokens for chat, code, speech, and OCR), tiered Vibe subscriptions (Free, Pro at $14.99/month, Team at $24.99/user/month, Education at $5.99/month, Enterprise quote-based), custom-priced enterprise deployments and Forge custom-model training engagements, and a managed Mistral Compute GPU cloud offering. The company reports approximately $400 million in annualized revenue at end of 2025, a 20x year-over-year increase, and is targeting over €1 billion in 2026, supported by 18,000 deployed NVIDIA Blackwell GPUs and a target of 200 MW of EU sovereign compute capacity by end of 2027.

Capital structure: a €105M seed in June 2023, a €414M Series A in December 2023, a €503M Series B in June 2024, a €1.7B Series C in September 2025 led by ASML at an €11.7B post-money valuation, and $830M of debt financing in March 2026 from a seven-bank consortium, with a further €3B round reportedly in talks in June 2026 at an approximately €20B valuation. Mistral has made two acquisitions in 2026 (Koyeb for serverless infrastructure in February, Emmi AI for physics simulation in May), holds partnerships with NVIDIA, SAP, IBM, Snowflake, TCS, Helsing, and major cloud providers, and has begun commercial expansion into Asia-Pacific with a Seoul sales presence.

Name

Mistral AI

Legal name

Mistral AI SAS

Website

https://mistral.ai

Company type

Private

Founded year

2023

Operating status

Operating

Headcount range

1,001–5,000 employees

Short description

Mistral AI is a Paris-based foundation model company that builds and deploys proprietary large language, code, speech, and document AI models, exposed through the Vibe agent, Studio, Forge, and Mistral Compute sovereign cloud platform for European enterprises, governments, and industrial customers.

Ownership category

akta.pro rank

Product category

Enterprise AI Platform

NAICS

Software Publishers (5132), Computer Systems Design and Related Services (54151)

SIC

Services-Computer Programming Services (7371)

akta.pro primary industry

Model Serving, Inference & Deployment Platforms (APIs, Edge/On-Prem) (HDAEANAC)

akta.pro secondary industries

Agents & Autonomous Workflows (Tool Use, Planning, Multi-Agent) (HDAAACAF), Model Deployment, Serving & Inference Platforms (HDAAABAF), AI/ML Engineering & Model Development Services (BPAEAHAF), Model Development & Training Platforms (AutoML, Notebooks, Feature Stores) (HDAEANAB)

Keywords

• Large language models
• Generative AI platform
• AI agent platform
• Sovereign AI infrastructure
• Custom AI model training

Headquarters

HQ city

Paris

HQ country

France

HQ region

Europe

Offices6 records

Markets served

GTM type

B2B

Offering type

Software

Cost components

Technology or R&D, Infrastructure, Personnel, Operations, Marketing or Sales

Revenue model

1. API and Usage-Based Model Access: Consumption-based revenue from token usage of Mistral models (Mistral Medium 3.5, Large 3, Small 4, Codestral, Voxtral, OCR 4, etc.) priced per million input/output tokens, with a 50% Batch API discount. Tool calls (code execution, web search, image generation, libraries indexing) are also billed per use.
2. Vibe Subscription Plans: Recurring subscription tiers for the Vibe consumer/SMB product: Free (limited), Pro at $14.99/month, Team at $24.99/user/month (minimum 50 users), and an Education plan at $5.99/month for verified students. Pro and Team include higher message, search, and image generation allowances plus coding sessions.
3. Enterprise Deployments and Custom Model Training: Custom-priced enterprise contracts covering self-hosted/private cloud deployments, custom models and agents via Forge, audit logs, SAML SSO, white-label, dedicated Mistral Compute capacity, and ongoing model customization and applied AI services.
4. Class: One-off training fees plus recurring monthly storage fees for fine-tuned models used for moderation, sentiment analysis, fraud detection, and similar classification tasks.
5. Mistral Compute (GPU Cloud): Managed dedicated GPU cluster capacity on Mistral-operated bare-metal infrastructure, sold to enterprises and AI labs requiring sovereign, EU-hosted compute for training and inference.
6. Cloud Partner Resale / Marketplace Distribution: Revenue share from Mistral models sold and consumed through cloud partner marketplaces and integrations (Google Cloud, AWS, Azure, SAP, IBM, Snowflake, NVIDIA, Outscale) and partner-led deployments (e.g., TCS as a Global System Integrator for Mistral Forge).

Pricing tiers

Go-to-market motion6 records

Marketing channels10 records

Core offering

Mistral AI develops and sells frontier large language models and a full-stack enterprise AI platform. Its product portfolio includes proprietary and open-weight foundation models (Mistral Large 3, Medium 3.5, Small 4, Codestral, Devstral, Voxtral, Mixtral, Magistral, Ministral), specialized models such as Mistral OCR 4 for document intelligence, the Vibe agent platform for long-horizon work, Mistral Studio for building agentic systems, Mistral Forge for custom model training, and Mistral Compute for sovereign GPU cloud infrastructure. The company sells these primarily to enterprises and governments in finance, public sector, and manufacturing through API access, on-premises deployments, and cloud marketplace distribution.

Product overview

Mistral AI offers a unified platform-plus-modules architecture of frontier AI products. The core product is Vibe (formerly Le Chat), Mistral's AI agent for long-horizon work, which combines a chat assistant (Vibe) and a coding agent (Vibe for Code) in a single product. Vibe is complemented by Mistral Studio, the AI production platform for building, deploying, and governing agentic AI systems; Mistral Forge, for training, aligning, and evaluating custom AI models; and Mistral Compute, frontier-scale GPU cloud infrastructure for training and inference. The platform also includes Mistral Search Toolkit for RAG and enterprise search, Physics AI (NeuralWing, NeuralMould) for industrial engineering optimization, and a family of foundation models such as Mistral Large 3, Mistral Medium 3.5, Mistral Small 4, Devstral 2, Codestral, Ministral 3, Voxtral (TTS, Mini Transcribe 2, Realtime, Small), Magistral, Mixtral 8x7B, and Mixtral 8x22B, plus Mistral OCR 4 for document intelligence. Applied AI services and custom model training provide expert-led deployment across serverless, dedicated, private cloud, and on-prem/edge environments.

Differentiator

Problem solved

Functional benefit

Brands

• Vibe: Consumer AI assistant and chatbot, originally launched as Le Chat and rebranded to Vibe in May 2026. Available on web, iOS, and Android.
• Le Chat
• Studio
• Forge
• Vibe for code
• Compute
• Voxtral
• OCR 4

Products and services

• Vibe (formerly Le Chat) Mistral's unified AI agent platform for long-horizon work combining a chat assistant (Vibe) and a coding agent (Vibe for Code) for individuals, teams, and enterprises, with 100+ tool connectors and full MCP compatibility.
• Vibe for Code Terminal-native, IDE-native, and web-based AI coding agent that writes, tests, and deploys code autonomously with full codebase context, powered by Devstral and Mistral Medium, with extensions for VS Code, JetBrains, and Zed.
• Mistral Studio Enterprise AI production platform to build, deploy, and govern agentic AI systems with workflows, agents, connectors, observability, guardrails, moderation, and a unified AI registry, deployable as self-hosted, on Mistral Cloud, or via cloud providers.
• Mistral Forge Custom model training and alignment platform supporting end-to-end training, SFT, DPO, RLHF, LoRA, and distillation, with synthetic data generation, drift detection, versioning, and rollback for domain-specialized LLMs.
• Mistral Compute Frontier-scale GPU cloud infrastructure for training and inference, providing dedicated hardware access (NVIDIA GB200, GB300, B300, Grace, x86 CPU nodes) on bare-metal with managed Kubernetes, managed Slurm, observability, SSO/SCIM, RBAC, and EVPN-VXLAN isolation. Target capacity is 200 MW sovereign across Europe by 2027.
• Mistral OCR 4 Fourth-generation document intelligence model that returns structured extraction with paragraph-level bounding boxes, typed block classifications (titles, tables, equations, signatures), and per-word confidence scores across 170 languages, deployable as a single container.
• Mistral Foundation Models (Large 3, Medium 3.5, Small 4, Devstral 2, Codestral, Voxtral, Mixtral, Magistral, Ministral 3)
• Voxtral Speech Stack (TTS, Mini Transcribe 2, Realtime, Small) Speech and audio AI suite including text-to-speech with voice cloning from 3-second samples, batch transcription with speaker diarization, real-time streaming transcription with sub-200ms latency, and speech understanding, supporting multiple languages.
• Applied AI Services (Custom Model Training)

Quantifiable outcome

• 99%+ OCR accuracy across 11+ languages on enterprise document benchmarks.
• +9 more outcomes

Named customers36 records

Segments8 records

Ideal customer profiles4 records

Technology focussed Yes

API detail

Has API

Yes

API docs

API detail

Core technology

AI maturity

App detail

Integration23 records

AI capability17 records

Feature9 records

Partnerships

19 partnerships are on record, tiered flagship, core and minor.

• NVIDIAflagshipTechnology or IntegrationMulti-pillar strategic partnership spanning Mistral's infrastructure (Mistral Compute) and models. The 44 MW Bruyères-le-Châtel data center is built on NVIDIA Grace Blackwell GB200 and GB300 systems, with early access to B300 and Grace/x86 CPU nodes. Jointly announced "Mistral Compute" as a sovereign AI platform on 18,000 Grace Blackwell Superchips. NVIDIA also participated in Mistral funding rounds.
• AirbusflagshipStrategic or Co-development PartnerStrategic partnership announced on May 28, 2026 covering the full Mistral AI product suite. Airbus acquired licences for on-premises deployment with guaranteed access to Mistral AI researchers and influence over the AI product roadmap. Collaboration spans engineering and design optimization, industrial operations automation, edge AI deployment on aircraft and spacecraft, and sovereign defence applications including cyber investigation and coding assistance.
• CMA CGMflagshipStrategic or Co-development PartnerStrategic partnership including MAIA (CMA CGM's internal AI assistant powered by Mistral). CMA CGM has also participated in Mistral's funding rounds, including the 2023 fundraising. Deployment spans shipping, logistics, and customer service across 160 countries for 155,000+ employees.
• HelsingcoreStrategic or Co-development PartnerStrategic partnership to develop next-generation AI systems for European defence, focused on Vision-Language-Action models that enhance defence platforms' ability to understand their environment, communicate with operators, and make rapid, reliable decisions in complex scenarios.
• French Agency for AI in Defense (Ministère des Armées)flagshipStrategic or Co-development PartnerStrategic partnership encompassing advanced research themes such as multimodal models, robotics, automation, and embedded systems, plus industrialization of key products for French defense.
• SAPcoreGTM or Marketing PartnerEuropean AI partner showcased at SAP Sapphire Madrid; Mistral is available through SAP's EU AI Cloud, which provides frontier AI models served entirely from European data centers as part of SAP's Autonomous Enterprise strategy. Includes Joule Studio for visual AI workflow orchestration.
• Tata Consultancy Services (TCS)coreImplementation/ SI/ Consulting PartnerTCS is the first Global System Integrator (GSI) to bring Mistral Forge to enterprises worldwide, offering frontier-grade AI solutions built on Mistral's customization platform.
• MicrosoftcoreTechnology or IntegrationOCR 4 and Document AI are available on Microsoft Foundry, and Mistral models are accessible via Microsoft Azure. Microsoft VP of AI Ecosystem Partnerships Kimmi Grewal highlighted the partnership as enabling advanced, structured document understanding directly into enterprise AI workflows.
• AWS (Amazon Web Services)coreTechnology or IntegrationMistral models are available on Amazon SageMaker, allowing customers to use Mistral via their AWS relationship and cloud credits.
• Google CloudcoreTechnology or IntegrationMistral models are available through Google Cloud, enabling joint customer deployments and use of Google Cloud credits for Mistral consumption.
• KoyebcoreStrategic or Co-development PartnerMistral AI announced the acquisition of Koyeb, a cloud computing startup based in Boulogne-Billancourt near Paris, on February 17, 2026 — Mistral's first acquisition. Koyeb's 13 employees and three co-founders joined Mistral, and the Koyeb serverless platform is being integrated into Mistral Compute to accelerate AI infrastructure development in Europe.
• Emmi AIcoreStrategic or Co-development PartnerMistral acquired Emmi AI to develop physics AI products for industrial engineers, including custom AI models and simulation tools for product design testing, feeding the Mistral physics AI initiative.
• VAST DatacoreTechnology or IntegrationMistral selected VAST Data's storage and AI Operating System stack software as the unified data platform layer for Mistral Compute's customer-facing cloud services and Mistral AI's internal model development and operations.
• MGX (Abu Dhabi)coreStrategic or Co-development PartnerMGX has partnered with France's Bpifrance, Mistral AI, and NVIDIA to develop Europe's largest AI campus in France with 1.4 gigawatts of compute capacity. Also referenced as a backer in the 50-billion-euro "AI Campus" data centre planned for Fouju, France.
• European Tech Creators coalition (ASML, Ericsson, Mistral, Nokia, SAP, Siemens)coreStrategic or Co-development PartnerMistral is a member of the European Tech Creators coalition alongside ASML, Ericsson, Nokia, SAP, and Siemens, and a founding member of the European Sovereign Tech Industry Alliance (ESTIA), advocating for European digital industrial policy and sovereignty.
• MozillacoreOEM/ Whitelabel/ Licensing PartnerMozilla's Firefox for Android "Shake to Summarize" feature (Firefox 152, June 16, 2026) uses Mistral Small (Mistral Small 3.1) cloud-based AI to generate page summaries, extending Mistral's reach to hundreds of millions of Firefox users.
• OnlyOfficeminorOEM/ Whitelabel/ Licensing PartnerOnlyOffice version 9.4 integrated Mistral AI natively into iOS mobile applications (extending the desktop plugin available since early 2025), enabling AI-powered content generation, reformulation, and task automation within document and spreadsheet editors.
• John Wiley & SonsminorStrategic or Co-development PartnerWiley's AI licensing business, which has generated nearly $100 million in revenue since early 2024, includes a partnership with Mistral AI (alongside Anthropic, AWS, and Perplexity) for content licensing.
• TotalEnergiescoreStrategic or Co-development PartnerTotalEnergies has a collaboration with Mistral AI to optimize operations and support AI-driven methane monitoring (MethaneLive), part of broader AI integration through data partnerships with Emerson and Cognite.

Scale indicators16 records

Recent moves11 records

Expansion highlights7 records

Direct peers

• OpenAI: Frontier large language model developer offering GPT-class models, ChatGPT consumer assistant, and enterprise APIs. Direct competitor to Mistral across frontier LLMs, coding agents, enterprise APIs, and sovereign/deployable offerings, with materially larger scale and revenue.
• Anthropic: Frontier AI lab behind the Claude model family, focused on enterprise and developer deployments with strong emphasis on safety and coding. Closely comparable to Mistral in enterprise go-to-market, coding models, and self-hosted/deployable offerings.
• Cohere: Enterprise-focused LLM provider offering customizable models, retrieval, and private deployments for regulated industries. Closely comparable to Mistral in its enterprise, customization, and data-residency positioning.
• xAI: Frontier AI lab building Grok models on its own large GPU clusters. Comparable to Mistral as a frontier-model competitor with proprietary infrastructure, though xAI's distribution is anchored in X/social rather than European enterprise.
• AI21 Labs: Enterprise foundation model company offering the Jamba and Task-Specific Models with an emphasis on customizability and on-premises deployment. Directly comparable to Mistral in offering private, customizable LLMs for enterprise customers.
• Aleph Alpha: German sovereign AI provider offering the Pharia model family with European data residency and on-premises deployments. Most direct European sovereign-AI peer to Mistral, competing for the same EU-regulated, sovereignty-minded workloads.

Broad incumbents

• Google DeepMind: Frontier AI research organization behind the Gemini family, integrated across Google Cloud and Workspace. Overlaps with Mistral in models, agents, and developer APIs, but operates as part of a much broader portfolio and via Google Cloud rather than a sovereign stack.
• Meta AI: Developer of the Llama family of open-weight foundation models. Direct peer on open-weight LLM strategy and competitive overlap on the cost-efficient end of the model spectrum, though Meta AI is a division of Meta rather than a standalone frontier vendor.

Emerging players

• Hugging Face: Open-source AI platform and model hub with growing enterprise and inference offerings. Overlaps with Mistral on open-weight model distribution, developer ecosystem, and enterprise AI tooling, though it does not own frontier model IP at Mistral's scale.
• Poolside AI: AI company focused on foundation models for software engineering. Overlaps with Mistral's Devstral, Codestral, and Vibe for Code on AI coding assistants, and competes for enterprise dev-tool spend, though it is narrower in scope.

Market position

Strengths5 records

Weaknesses5 records

Competitive moat8 records

Key risks6 records

Key highlights7 records

Customer concentration

• Xx.com/aimistral
• GitHubgithub.com/mistralai
• Facebookfacebook.com/mistralai
• LinkedInlinkedin.com/company/mistralai
• Instagraminstagram.com/mistral.ai
• YouTubeyoutube.com/@mistralai

Revenue estimate

Valuation estimate

Number of profiles

Profiles7 records

Subsidiaries2 records

Funding overview

Funding rounds10 records

Investors56 records

Funding detail is available on the Subscription and Enterprise plan.Contact sales →

M&A3 records

Investments1 record

M&A and investment is available on the Subscription and Enterprise plan.Contact sales →

Frequently asked questions about Mistral AI