Mistral AI Statistics (2026): €21B+ Valuation and 61 Key Stats

A data roundup of Mistral AI’s funding, reach, Le Chat assistant, search products, models and environmental footprint, taken only from Mistral’s own announcements and pricing page plus the Hugging Face and GitHub APIs.

Mistral AI is Europe’s best-funded AI lab and one of the most downloaded open-weight model makers, but it is not where most buyers ask AI for product advice. Mistral raised €3 billion at a post-money valuation of more than €21 billion (Series D), and Mistral-7B-Instruct-v0.2 alone has 66,301,807 all-time downloads on Hugging Face. Its assistant (Le Chat, now part of Mistral’s Vibe apps) has web search, AFP news and a $14.99 Pro plan (pricing), yet Mistral publishes no user count. Pick Mistral data for enterprise, sovereignty and open-model stories; skip it as a proxy for where your customers research software. For that, measure ChatGPT, Perplexity, Claude, Gemini and Google AI answers with AI visibility tools.

Key takeaways

• €3 billion Series D at a post-money valuation of more than €21 billion (Series D)
• 66,301,807 all-time Hugging Face downloads for Mistral-7B-Instruct-v0.2 (Hugging Face)
• 125+ global enterprises supported, across 20 countries (Series D)
• 1 trillion parameters in Mistral Large 4, with 52 billion active (Mistral Large 4)
• $14.99 a month for Mistral Pro, or $5.99 for verified students (pricing)
• 86% correctness on FinanceBench for Agentic Search, up from 26.7% (Agentic Search)

How many people use Mistral AI?

Mistral does not publish monthly active users for its assistant, so the best public signals are open-model downloads and developer attention.

1. Mistral-7B-Instruct-v0.2 has 66,301,807 all-time downloads on Hugging Face (Hugging Face API).
2. Mistral-7B-Instruct-v0.2 was downloaded 1,831,711 times in Hugging Face’s latest monthly count (Hugging Face API).
3. Mistral-7B-Instruct-v0.3 has 48,634,386 all-time downloads (Hugging Face API).
4. The original Mistral-7B-v0.1 base model has 27,948,575 all-time downloads (Hugging Face API).
5. Mixtral-8x7B-Instruct-v0.1 has 24,067,252 all-time downloads (Hugging Face API).
6. Mixtral-8x7B-Instruct-v0.1 has 4,776 likes on Hugging Face (Hugging Face API).
7. Voxtral-Mini-4B-Realtime-2602, Mistral’s speech model, has 13,918,554 all-time downloads (Hugging Face API).
8. Voxtral-Mini-4B-Realtime-2602 was downloaded 2,033,798 times in the latest monthly count, more than any 7B model (Hugging Face API).
9. Mistral-Nemo-Instruct-2407 has 10,643,531 all-time downloads (Hugging Face API).
10. Mistral-Small-3.1-24B-Instruct-2503 has 4,689,444 all-time downloads (Hugging Face API).
11. The mistralai GitHub organization has 9,281 followers (GitHub API).
12. Mistral has 31 public repositories on GitHub (GitHub API).
13. The mistral-inference repository has 10,821 stars (GitHub API).
14. mistral-inference has been forked 1,063 times (GitHub API).
15. The mistral-vibe coding agent repository has 5,213 stars (GitHub API).

How much is Mistral AI worth?

16. Mistral raised €3 billion in a Series D funding round (Series D).
17. The Series D valued Mistral at more than €21 billion post-money, which Mistral calls the largest equity round ever completed by a European technology company (Series D).
18. Mistral’s Series C raised €1.7 billion in September 2025 (Series C).
19. The Series C valued Mistral at €11.7 billion post-money (Series C).
20. Mistral now operates across 20 countries (Series D).
21. Mistral supports 125+ global enterprises, including Airbus, ASML and HSBC (Series D).
22. Mistral was founded in April 2023 (About).
23. Mistral closed its seed round on June 13, 2023 (About).
24. Mistral hired its 100th employee on September 30, 2024 (About).
25. Mistral plans to build up to 1 GW of compute capacity by 2030 (regional inference).
26. Its new Les Ulis site is a 10 MW facility dedicated to inference (AI Now Summit).

Le Chat and Mistral search statistics

Le Chat is Mistral’s assistant; its pricing page now sells it inside the Vibe apps. These are the figures that matter for AI search.

27. Mistral Pro costs $14.99 a month (pricing).
28. Verified students pay $5.99 a month for Pro (pricing).
29. Team costs $24.99 per user per month (pricing).
30. Pro and Team allow up to 5x the Free plan’s web searches (pricing).
31. Pro allows up to 40x more image generations than Free (pricing).
32. Team gets 200 Flash answers a day, against 150 on Pro (pricing).
33. Le Chat’s Flash Answers respond at up to ~1000 words per second (Le Chat launch).
34. Students got over 50% off Le Chat Pro at its February 2025 relaunch (Le Chat launch).
35. Le Chat’s news answers draw on AFP’s 1,700 journalists worldwide (AFP partnership).
36. Le Chat uses AFP’s daily output of 2,300 wires in six languages (AFP partnership).
37. Mistral’s Agentic Search raised correctness on financial filings from 26.7% to 86% on FinanceBench (Agentic Search).
38. On OfficeQA Pro, Agentic Search went from 6.3% to 51.9%, a 45.6-point gain (Agentic Search).
39. Agentic Search cuts p90 latency by up to 39.6% (Agentic Search).
40. Fewer repeated searches cut token use by up to one-third (Agentic Search).
41. A Search Toolkit pipeline built with CMA CGM returns fake-news alerts within 15 seconds end to end (Search Toolkit).

Mistral model statistics

42. Mistral Large 4 is a 1 trillion-parameter model with 52 billion active parameters (Mistral Large 4).
43. Large 4 was trained on 3,800 NVIDIA Grace Blackwell GPUs in Mistral’s own European datacenters (Mistral Large 4).
44. Large 4’s training data spans more than 160 languages (Mistral Large 4).
45. Large 4 scores 61.7% on DeepSWE v1.1, 59.4% on SWE-Atlas-QnA and 28.3% on Terminal-Bench 4 (Mistral Large 4).
46. Large 4 scores 59.9% on AutomationBench, a set of 657 business workflows (Mistral Large 4).
47. Mistral Large 3 has 675B total and 41B active parameters (Mistral 3).
48. Large 3 was trained on 3000 NVIDIA H200 GPUs (Mistral 3).
49. Ministral 3 14B reasoning scores 85% on AIME ‘25 (Mistral 3).
50. Mistral Small 4 has 119B total parameters, 6B active per token (Mistral Small 4).
51. Small 4 cuts end-to-end completion time by 40% in a latency-optimized setup (Mistral Small 4).
52. Small 4 has a 256k context window (Mistral Small 4).
53. Mistral Medium 3 performs at or above 90% of Claude Sonnet 3.7 on benchmarks at $0.4 input and $2 output per million tokens (Mistral Medium 3).
54. Mistral 7B outperformed Llama 2 13B on all metrics at launch in September 2023 (Mistral 7B).
55. Mistral Large API costs $0.5 per million input tokens and $1.5 per million output tokens (pricing).
56. Batch processing cuts API prices by 50% (pricing).
57. Cached input tokens cut input cost by up to 90% for repeated prompts (pricing).

Source: Mistral’s Large 4, Mistral 3 and Small 4 announcements, checked October 27, 2026.

What is Mistral AI’s environmental footprint?

Mistral ran a lifecycle analysis of Mistral Large 2 with Carbone 4 and the French agency ADEME.

58. Training and 18 months of use of Mistral Large 2 emitted 20,4 ktCO₂e (environmental study).
59. Mistral Large 2 consumed 281 000 m3 of water over the same period (environmental study).
60. One 400-token Le Chat response emits 1.14 gCO₂e (environmental study).
61. One 400-token Le Chat response uses 45 mL of water (environmental study).

What to do with this data

• Don’t read downloads as buyers. Sixty-six million downloads of one 7B model are developers running Mistral themselves. They show Mistral matters in the model supply chain, not that your buyers ask Le Chat which product to buy.
• Mind the enterprise layer. Mistral sells Agentic Search, Search Toolkit and private deployments to 125+ enterprises. Clear, crawlable, well-structured pages are what any retrieval layer, Mistral’s included, can find and cite. Our AI crawler statistics cover the bots doing that work.
• Measure the engines that send pipeline first. Mentionova tracks ChatGPT, Perplexity, Claude, Gemini, Google AI and Reddit, and its agents draft the briefs, content and Reddit replies that close citation gaps. It is #1 in our best AI visibility tools and LLM brand monitoring tools roundups.
• Compare the engines. Read these numbers next to our ChatGPT statistics, DeepSeek statistics, Grok statistics and AI chatbot statistics.

Frequently asked questions

How much is Mistral AI worth?

Mistral’s Series D raised €3 billion at a post-money valuation of more than €21 billion. Its Series C in September 2025 raised €1.7 billion at €11.7 billion.

How many people use Mistral AI?

Mistral does not publish user counts for Le Chat. Its models are widely used by developers: Mistral-7B-Instruct-v0.2 has 66,301,807 all-time downloads on Hugging Face, and Mistral supports 125+ global enterprises.

Is Le Chat free?

Yes. Mistral’s Free plan gives limited access to its latest models, chat, search, creation and Vibe for code. Pro is $14.99 a month ($5.99 for verified students) and Team is $24.99 per user per month.

Does Le Chat search the web?

Yes. Every plan includes web searches, with up to 5x the Free allowance on Pro and Team, and Le Chat grounds news answers in AFP’s 2,300 daily wires.

How big is Mistral Large 4?

Mistral Large 4 has 1 trillion parameters with 52 billion active, and was trained on 3,800 NVIDIA Grace Blackwell GPUs in Mistral’s European datacenters.

How we compiled these statistics

We fetched 26 primary source pages on October 27, 2026 and recorded 61 data points in docs/content/stats-ledgers/mistral-ai-statistics.json. Every numbered figure above maps to a ledger row with a verbatim quote. Funding, company, product, pricing, model and environmental figures are Mistral’s own statements from its news posts, About page and pricing page. Download, like, star and follower counts come straight from the Hugging Face and GitHub APIs on the fetch day and will keep rising. Benchmark claims in Mistral’s announcements are self-reported. We did not use market-research sites or traffic estimators. Related reading: AI search statistics, Perplexity statistics, Microsoft Copilot statistics.