Skip to content

Mistral AI

Dossier checked 28 Sep 2026

What it sells now

Newest frontier release

Mistral Large 4 (open weights)

29 Oct 2026 · the lab’s announcement

Improved Mistral Large 4 after three more weeks of reinforcement learning, released as open weights under the Mistral Ecosystem Open-Weight licence; list price set at $0.70 in and $2.10 out

Flagship list price, per million tokens

$1.50 in $7.50 out

Mistral Medium 3.5, its first flagship merged model, released 28 Apr 2026 · pricing page

Mistral Large 4 entered public preview on 6 October; on 29 October Mistral released an improved checkpoint as open weights and set its list price at $0.70 in and $2.10 out. Medium 3.5 is priced here as the generally available model.

Weights

Open flagship

Mostly open. Medium 3.5 under a modified MIT licence with an exception for companies with large revenue; Large 3, Ministral 3, Small 4 and Shieldstral under Apache 2.0; Voxtral TTS under CC BY-NC 4.0. OCR, Codestral and the embedding models stay closed. Source

Where you can buy the flagship

Its other models, not the flagship

Source

8 releases in the year to 30 Oct 2026, 8 the year before · 24 in all since 2023

Our read, not a fact

In the model

The question we ask of every lab: is most of its value in the model, which gets cheaper fastest, or in something other work is built on? Our answer: in the model itself.

Mistral ships open weights, the fastest-commoditising part of the stack. The durable position is European sovereignty, on-prem enterprise trust, and Le Chat as a distribution surface.

What would change our readOpen weights stop tracking the closed frontier, or sovereignty stops being something European buyers will pay for.

01 · What it runs on, and the money

Trains and serves on

Its announced compute is NVIDIA: Mistral Compute, its own cloud, targets up to 1 GW by 2030, and an August 2026 deal with HUMAIN explores capacity in Saudi Arabia. Source 1 Source 2

What it depends on Every disclosed buildout uses NVIDIA silicon, so its sovereignty covers ownership, siting and jurisdiction, not chip supply. Source

Biggest risk Its whole disclosed compute base is NVIDIA silicon, and NVIDIA is also an investor. Source

Raised
EUR 3B Series D (Sep 2026), after a EUR 1.7B Series C (Sep 2025), plus $830M of debt (Mar 2026) Source
Valuation
More than EUR 21B post-money (Series D, 8 Sep 2026, led by Samsung Electronics) Source
Structure
Mistral AI SAS, a French société par actions simplifiée. Private, unlisted, founded April 2023 in Paris. No dual-class or foundation structure is documented. Source

NotesThe year-ago comparison is Mistral Medium 3 at its May 2025 launch price. Large 3 is still on sale at $0.50 in and $1.50 out. Revenue and headcount are not published. An open-weight frontier model trailed for summer 2026 had not shipped by 28 Sep 2026.

02 · Releases

data through 30 Oct 2026

  1. Mistral 17 Mistral Large 4 (open weights)29 Oct 2026

Each dot is a frontier release at its date. Click one to find it in the full log. Every lab on one chart

The landmarks

  1. 2 Dec 2025 Mistral Large 3 Mixture-of-experts flagship, open weights and on the API from launch MoEFlagship
  2. 10 Jun 2025 Magistral First reasoning models: Magistral Medium in public preview on the API and Le Chat, Magistral Small as open weights ReasoningFirst of kind
  3. Dec 2023 Mixtral 8x7B Open mixture-of-experts breakthrough OpenMoE
  4. Sep 2023 Mistral 7B First model, punched far above its size OpenFirst model
The full release logall 24, newest first
  1. 2026
  2. 29 Oct 2026 Mistral Large 4 (open weights) Frontier Improved Mistral Large 4 after three more weeks of reinforcement learning, released as open weights under the Mistral Ecosystem Open-Weight licence; list price set at $0.70 in and $2.10 out Source
  3. 6 Oct 2026 Mistral Large 4 Frontier Flagship in public preview on the API from 6 Oct; an improved checkpoint followed as open weights on 29 Oct Source
  4. Jul 2026 Robostral Navigate First embodied-navigation model
  5. Jul 2026 Leanstral 1.5 Formal proof and verified coding
  6. Jun 2026 Mistral OCR 4 Document intelligence
  7. 28 Apr 2026 Mistral Medium 3.5 Frontier Mid tier, powering remote coding agents; on the API from 28 Apr, launch post 22 May Source
  8. Mar 2026 Mistral Small 4 Small tier refresh
  9. 2025
  10. 2 Dec 2025 Mistral Large 3 Frontier Mixture-of-experts flagship, open weights and on the API from launch Source
  11. 17 Sep 2025 Magistral Medium 1.2 Frontier Multimodal update of the Medium-tier reasoning model, on the API Source
  12. 12 Aug 2025 Mistral Medium 3.1 Frontier Medium-tier update on the API, improving tone and performance Source
  13. 24 Jul 2025 Magistral Medium 1.1 Frontier Update of the Medium-tier reasoning model, on the API Source
  14. 10 Jun 2025 Magistral Frontier First reasoning models: Magistral Medium in public preview on the API and Le Chat, Magistral Small as open weights Source
  15. 7 May 2025 Mistral Medium 3 Frontier Mid-tier flagship, on the API from launch Source
  16. 17 Mar 2025 Mistral Small 3.1 Frontier Open multimodal Small update; Mistral’s second tier while no Medium was on sale Source
  17. 30 Jan 2025 Mistral Small 3 Frontier Efficient open model; Mistral’s second tier while no Medium was on sale Source
  18. 2024
  19. 18 Nov 2024 Mistral Large 2.1 Frontier Upgraded Large 2 on the API and Le Chat, open weights under the Mistral Research License Source
  20. Sep 2024 Pixtral First multimodal model
  21. Jul 2024 Mistral Large 2 Frontier 123B-parameter flagship
  22. May 2024 Codestral 22B Dedicated code model
  23. Apr 2024 Mixtral 8x22B Larger open MoE
  24. Feb 2024 Mistral Large Frontier First proprietary flagship
  25. Jan 2024 Mistral Medium Frontier La Plateforme’s flagship endpoint: announced 11 Dec 2023 behind a waitlist, open to anyone who subscribed by 20 Jan 2024
  26. 2023
  27. Dec 2023 Mixtral 8x7B Frontier Open mixture-of-experts breakthrough
  28. Sep 2023 Mistral 7B Frontier First model, punched far above its size

03 · When to reach for them

practical read

Open weights to self-host
Reach for Mistral Large 3 when you need Apache 2.0 weights to self-host or fine-tune (Dec 2025).
On-prem and sovereignty
Pick Mistral Medium 3.5 when data must stay on your own hardware; its weights are downloadable, with a licence exception for large-revenue firms (May 2026).
Low-cost open model
Use Mistral Small 4 for Apache 2.0 work at $0.15 in and $0.60 out per million tokens (Mar 2026).

04 · Weights, across all ten

the flagship

Open flagship
DeepSeek Mistral AI Alibaba (Qwen) Moonshot AI Zhipu (Z.ai)
Closed flagship, some open
OpenAI Google Meta xAI
Closed
Anthropic

05 · Essays that read Mistral AI

from the archive

The Model Is Not the Moat If frontier capability keeps centralising, the durable edge shifts outward into trust, workflow fit, and the surrounding package. The Forest Floor Is The Product Most people are optimising for canopy. The work that survives the next five AI releases is built from the forest floor up. You're Not Comparing Models. You're Comparing Contracts. Agent benchmarks don't measure models. They measure contracts. Two teams running the same model can publish different scores, and both can be honest.

06 · About Mistral AI

at a glance

Founded
2023
HQ
Paris, France
Leadership
Arthur Mensch (CEO)
Founders
Mensch, Lample, Lacroix
Also builds
Le Chat, Mistral Studio, Mistral Compute
Backers
Samsung, ASML, a16z, NVIDIA

Founded April 2023 by Arthur Mensch, Guillaume Lample and Timothée Lacroix, former DeepMind and Meta researchers. Headquartered in Paris, it is Europe’s leading frontier lab, known for high-quality open-weight models alongside proprietary ones and its assistant, Le Chat.