At a glance
- Category
- AI model providers
- Tags
- EU-hosted
Company
- Founded
- 2023
For developers
- SDKs
- Python, TypeScript
- Runs
- Cloud, Self-hosted, On-prem, Edge
- Works with
- BNP Paribas, AXA, Synapse Medicine, Pierre Fabre, CMA CGM, Zalando and 17 more
Security
- Compliance
- SOC 2 Type II, ISO 27001, ISO 27701
What is Mistral AI?
Mistral AI is an enterprise AI platform for teams that build, deploy, and operate tailored AI systems across the full lifecycle. Studio combines Agent Runtime, Observability, AI Registry, datasets, and experiments, and Vibe (renamed from Le Chat in May 2026) is the single chat and coding assistant for research, writing, and agentic coding. It connects with GitHub, GitLab, Jira, Snowflake, Kubernetes, and SAP, and customer stories include Stellantis, ASML, and CMA CGM.
The 2026 rebrand: Le Chat is now Vibe
On 28 May 2026 Mistral folded its chatbot Le Chat into a single product called Vibe, available in three modes: Work (research, writing, and multimodal tasks on web and mobile), Code (a CLI, a VS Code extension, and a managed cloud coding sandbox), and a legacy Chat mode that preserves old Le Chat workflows. The developer console that used to be called La Plateforme is now branded Studio. Source: Mistral's own changelog (docs.mistral.ai/resources/changelogs, 28 May 2026 entry) and the current Vibe product page, whose own meta description reads "Vibe (formerly Le Chat)". Anyone following older guides, blog posts, or this listing's own prior copy will still see the old names; the underlying product and pricing didn't change, only the branding and the fact that chat and coding now live in one app instead of two.
Pricing: clear for individuals, opaque for Enterprise
Vibe has four tiers, confirmed on mistral.ai/pricing (checked 12 October 2026): Free (limited messages, web search, coding sessions, and image generation); Pro at $14.99/month (higher limits, full Vibe coding access; verified students pay $5.99/month); Team at $24.99 per user/month (shared workspace, admin controls, up to 30GB storage per user, domain verification, data export); and Enterprise, which is contact-sales only (custom models, custom agents/workflows, audit logs, SAML SSO, white-labeling). API pricing is also public and genuinely transparent: most models are billed per million tokens, with Mistral Large priced at $0.50/M input and $1.50/M output tokens as the page's own example; batch processing cuts the price 50%, and cached input tokens can cut input cost up to 90% on repeated prompts. OCR is billed per 1,000 pages, speech models per minute, and tool/agent calls per call.
Security and compliance: certifications you can actually check, and incidents they chose to disclose
Mistral runs a public Vanta-hosted trust center at trust.mistral.ai that lists, as of this check, an ISO/IEC 27001:2022 certificate, an ISO/IEC 27701:2019 certificate, a Statement of Applicability, a 2026 penetration testing report, and a SOC 2 Type II letter of attestation (the full SOC 2 report is enterprise-customers-only). That's a materially more checkable security posture than most AI vendors publish. The same page also discloses two real incidents rather than staying quiet: a 'Mini Shai-Hulud' supply-chain attack on one of its codebase management systems, disclosed 14 May 2026, which the company says was contained to 'certain non-core code repositories' with no hosted services, customer data, or research environments affected, and investigation subsequently closed; and six shell-permission CVEs in the Vibe CLI (CVE-2026-87983 through CVE-2026-87988), disclosed and patched the same day, 12-14 September 2026, in Vibe 2.25.4. Buyers giving Vibe's CLI shell/terminal access should keep it current given that history, but the pattern of prompt disclosure (plus two earlier security changelog entries from 2024-2025 covering a Le Chat prompt-injection fix and a token-length side-channel fix reported by Microsoft researchers) is a point in the vendor's favor, not a red flag by itself.
Self-hosting and licensing
Mistral's open-weight models (e.g. Mistral 7B and the Mixtral family) are Apache 2.0 licensed for research and individual use, per the vendor's own pricing FAQ, but commercial/production self-hosted deployment of those models requires a separate Mistral license with its own terms for derivatives. Some newer models ship under more permissive terms: Mistral Small 4 (March 2026) was released fully under Apache 2.0, and Mistral Medium 3.5 (April 2026) under a Modified MIT license, per the vendor's changelog — so licensing now varies by model rather than being uniform, and it's worth checking the specific model card before assuming terms. If you're planning to run weights yourself: the mistralai/mistral-inference reference library (the simplest documented path to running Mistral weights locally) was archived in June 2026 with no named successor; mistralai/client-python and mistralai/mistral-common, by contrast, were both pushed to within the last few days as of this check, so the API-client side of the org is clearly active even though that one self-hosting helper library isn't.
Company, funding, and why it keeps winning strategic partners
Mistral AI SAS was founded in Paris in April 2023 by Arthur Mensch (ex-Google DeepMind), Guillaume Lample, and Timothée Lacroix (both ex-Meta). Valuation has climbed from roughly €240M (about $264M) at its first raise (June 2023) to €2B (about $2.2B) (December 2023), €5.8B (about $6.4B) (June 2024), and €12B (about $13B)/$14B in September 2025 after a €2B (about $2.2B) round that included a €1.3 (about $1.43)-1.5B, 11%-equity stake from ASML — Europe's largest tech company by market cap. In September 2026, Samsung Electronics took a stake in a €3B (about $3.3B) transaction that the company says brought its valuation to roughly €21B (about $23B). Mistral also has active infrastructure and distribution partnerships with Microsoft (Mistral Medium 3.5 and OCR 4 now ship in Microsoft Foundry and Copilot Studio, July 2026), NVIDIA (open frontier model development, March 2026), and Accenture (enterprise deployment, February 2026). Source: en.wikipedia.org/wiki/Mistral_AI, corroborated against Mistral's own changelog for the product-side dates. These partnerships, plus the explicit 'digital sovereignty' positioning, are why Mistral shows up repeatedly in EU public-sector and regulated-industry deployments (its own site names ASML, Stellantis, CMA CGM, SNCF, BNP Paribas, La Banque Postale, Abanca, and the Government of Luxembourg as customers, which are vendor-attributed case studies, not independently audited figures).
What Studio, Vibe, Forge, and Compute actually do
Studio (console.mistral.ai) is the enterprise build/deploy/govern platform: agent and workflow orchestration, an AI registry with versioning and rollback, observability (traces, dashboards), guardrails and moderation, and deployment as hybrid, dedicated, or fully self-hosted. Vibe is the unified assistant described above. Forge is a separate product for organizations that want to train or fine-tune a model on their own data end-to-end (data prep, training, RLHF/LoRA/DPO alignment, KPI-based evaluation, and inference), evidenced on the vendor's own page by case-study logos for ASML, Ericsson, ESA, Reply, and DSO. Compute and Inference (under /cloud/) are the underlying serving and training infrastructure Mistral sells directly, pitched specifically as a European alternative to US hyperscaler compute. None of the usage numbers Mistral cites for these products (e.g., developer adoption or code-completion-accuracy percentages sometimes shown in its own marketing) are independently verified, and should be read as vendor claims.
Mistral AI price and terms
Terms
- Free plan
- YesYes
The terms in full
Standing Free plan with limited messages, web searches and coding sessions, basic image generation, up to 5 scheduled tasks, plus access to Mistral Studio for the API; usage beyond it is per-token API billing or a paid plan.

Pricing page, captured
mistral.ai/pricing, checked
All plans and limits
| Plan | Price | Main limits |
|---|---|---|
| Pro (verified students) | $5.99 a month | None stated |
| Pro | $14.99 a month | Flash answers: 150 / day |
| Team | $24.99 per user a month | Flash answers: 200 / day; Storage: Up to 30GB per user |
| API: Mistral Large 4 (input) | $0.70 per M tokens | None stated |
| API: Mistral Large 4 (output) | $2.10 per M tokens | None stated |
| API: Voice Agents (Mistral Small 4) | $0.04 per minute | None stated |
| Enterprise | Price on request | None stated |
Read from mistral.ai/pricing and mistral.ai/pricing/api.
What we checked on Mistral AI
What the price counts
Mixed metering: the API is billed per million input and output tokens (OCR per page, speech per minute, tools per call), while Vibe plans are a flat $14.99 monthly Pro, $24.99 per user Team, and quote-only Enterprise.
Most models are priced per million tokens, with both input (your prompts) and output (responses) counted separately.
mistral.ai/pricingChecked
Free to use
A standing Free plan gives limited access to Vibe chat, search, coding in the terminal and Mistral Studio, with message and session caps rather than a time limit.
Yes, Mistral’s Free plan provides limited access to Vibe, including our latest models, chat, search, creation, and Vibe for code for agentic coding…
mistral.ai/pricingChecked
Open source
18 public repositories on GitHub, none of them the product's code
On the page: github.com/mistralai repositories
github.com/mistralaiChecked by us,
Mistral AI alternatives
Mistral AI and Fireworks AI
Mistral AI beside its alternatives: the price on each one's site and the day it was checked Tool Price Checked Mistral AI Free, then $14.99 a month Fireworks AI From $0.05 per 1m input tokens
Questions about Mistral AI
- How much does Mistral AI cost?
- Mistral AI has a free plan. Its paid plans include Pro at $5.99 a month, Pro at $14.99 a month and Team at $24.99 per user a month. From mistral.ai/pricing, checked 18 November 2026.
- Does Mistral AI have an API or an MCP server?
- Mistral AI has an API.
- What have we checked on Mistral AI?
- 2 facts, under one heading: What it costs to start.
- How do we check Mistral AI?
- We hold 2 facts on Mistral AI, quoted from 3 pages. Facts and price last checked 18 November 2026, when we also found the last change.
Sources and changes
Last changed
Sources on mistral.ai3
What changed1
- What it does
Mistral AI price history
The cheapest paid plan has cost $14.99 a month since 3 November 2026, on all 3 readings.
| Price | Plan | First seen | Last seen | Readings | Source |
|---|---|---|---|---|---|
| $14.99 a month | 3 November 2026 | 19 November 2026 | 3 | Archived copy, 3 November 2026 |
From 8 readings of Mistral AI's pricing page: 6 Internet Archive copies from 25 May 2025 and 2 of our own checks since 18 November 2026. A date is when we read the page, not when the price changed.
Similar to Mistral AI
- ERIUS PHONERented cloud Android phones, each in its own container, that keep an agent's apps and sign-ins between sessions.
- TinyFishWeb search, page fetching, cloud browsers and goal-driven web automation for AI agents, with search and fetch free.
- SmitheryPublic MCP server registry, now run by Arcade, listing direct vendor MCP URLs after its proxy shut down.
- UI DiscoveryA library of real websites, iOS apps and dashboards with each one's palette, fonts and original build prompt.
From our research
- Best sales agents for marketing teams, and what to checkApollo suits teams starting free, VoiceDrop suits voicemail campaigns, and PipeLime suits hands-off email and LinkedIn prospecting.
- Which browser automation API should you build an agent on?Skyvern, Browserbase, Browser Use, Playwright MCP, Hyperbrowser, and TinyFish compared on credential handling, session isolation, and audits, not speed.
- Clay pricing: every plan, its Actions, Data Credits and rowsWhat Clay's Free, Launch, Growth and Enterprise plans cost, and how many Actions, Data Credits and table rows each one includes.
Is Mistral AI yours?
Free, and the page then carries the verified owner mark. You confirm it with an email at mistral.ai. Claim this listing
Report an error if a fact is wrong.