Mistral built its enterprise stack in eight days and priced none of it
Between October 29 and November 5 the Paris lab published its biggest model's weights, expanded its agent platform and bought a fifth startup, all funded by a September mark of €21 billion. Not one piece of the stack carries a published price, while free Chinese open-weight models still outscore it.
Vincent Jiangthree-minute read
Between October 29 and November 5, Mistral AI$14B — Mistral AI, private, latest valuation $14B published the downloadable weights of Mistral Large 4, the trillion-parameter model it calls Le Chonk, expanded its enterprise platform Mistral Studio with managed agents, workflows and hosted apps, and bought SkipLabs, a compiler startup founded by former Meta+0.68% — Meta, up 0.68 percent today engineer Julien Verlaguet, its fifth acquisition of 2026 123.
One item needs a correction: Mistral Compute, the private GPU infrastructure offering with ten named launch partners including Orange, SNCF and Thales, is dated June 11, 2025, on Mistral's own blog, so this week's coverage re-surfaces a year-old offer rather than a fresh launch 76.
The mark was set before the stack existed
None of this was cheap. In September Mistral raised €3 billion at a post-money valuation above €21 billion, nearly double last year's €11.7 billion, in a round led by Samsung with ASML+0.45% — ASML, up 0.45 percent today, Nvidia−4.65% — Nvidia, down 4.65 percent today and Salesforce−0.72% — Salesforce, down 0.72 percent as of its last quote returning and BlackRock−0.04% — BlackRock, down 0.04 percent as of its last quote funds and Luxembourg joining 5. The only revenue anchor is a company forecast that annual recurring revenue will pass $1 billion this year from more than 125 enterprise customers, a forecast, not audited revenue, and it was given before the agent platform, the on-premises push or the SkipLabs deal existed 5.
Nothing in the stack has a price
The announcements answer every question except the commercial one. Mistral published no pricing for Studio's managed agents, apps or workflows, which reach on-premises deployments in late November 2. Mistral Compute has no price, no self-serve signup and no public capacity timeline beyond its launch partners 67. The SkipLabs price was not disclosed; the startup had raised $8 million in seed funding in 2025 34.
The free alternative outscores the flagship
Here is the tension the pitch has to overcome. On independent firm Artificial Analysis' Intelligence Index v5, Large 4 scores 26, the best of any open-weight model developed outside China, but six points behind the open-weight leaders, China's DeepSeek V4.1 Flash 1017 and GLM-5.3, both at 32 1. On cost per benchmark task, Mistral charges $2.06 where DeepSeek measures $0.60 1. The same banks Mistral courts can run those weights free.
Le Chonk tops the non-Chinese open models, six points behind the leaders
Le Chonk leads the non-Chinese open models and trails the Chinese leaders by six points
Data
| Value | |
|---|---|
| DeepSeek V4.1 Flash 1017 | 32 |
| GLM-5.3 | 32 |
| MiMo-V2.6-Pro | 30 |
| Mistral Large 4 | 26 |
| Muse Spark Open | 25 |
Why sovereignty may still command a premium
The counter-argument is control. BNP Paribas says it runs Studio on its own infrastructure to keep sensitive processes in hand 2, and Mistral's chief revenue officer Marjorie Janiewicz sums up customer demand as "I want to own my model" 1. SkipLabs' six engineers join chief scientist Guillaume Lample's group to build verification into code-writing agents, trust that rented tools do not supply 4. Sovereignty is a real product. It just has no listed price against it yet.
Three tests before the price tags appear
Three dated tests settle it: on-premises Studio arriving late November, whether any price appears for agents or Compute, and whether Skip stays open source under its new owner 23. Until then, €21 billion is priced on a forecast 23. The mark came first; the price tags never did.
Get Model by email, free
The AI labs, their models, and the business behind them. In our daily email, every weekday.
