From Apple's acquisition rumors to ASML investing 1.3 billion to become a major shareholder, uncover the technological and business secrets of Mistral AI.
In early September, Apple was reported to be interested in acquiring French startup Mistral AI. Shortly after, semiconductor giant ASML led its Series C financing with 1.3 billion euros. Currently, the company's valuation has soared to $14 billion, becoming the hottest benchmark force in the European AI track.
The rapid changes in the technology industry have made every move of the giants highly anticipated. Recently, rumors about Apple's intention to acquire Mistral AI have been rife. On September 9th, Dutch lithography giant ASML officially announced that it would lead Mistral AI's Series C financing with 1.3 billion euros and establish a strategic partnership with it. These series of actions quickly made Mistral AI the focus of discussion. One can't help but wonder, what exactly is Mistral AI? Why can it attract the pursuit of technology giants? What unique advantages does it have to stand out in the highly competitive AI track?
Favored by Giants, Emerging and Drawing Attention
Mistral AI was officially founded in Paris, France in April 2023. Although it has a short history, it has quickly emerged in the AI field. This company was jointly founded by three post - 90s geniuses - Arthur Mensch, Timothée Lacroix, and Guillaume Lample. Its team is luxurious, and its core members are all from top institutions such as DeepMind and Meta.
Guillaume Lample, Arthur Mensch, Timothée Lacroix (from left to right), co - founders of Mistral AI
Since its establishment, Mistral AI has shown rapid development momentum. Just two months after its establishment and before releasing its first batch of models, it successfully raised 105 million euros (approximately $117 million), setting the largest seed - round financing record in European history at that time. Subsequently, in December 2023, it completed a Series A financing of 385 million euros ($415 million); in June 2024, it received a Series B financing of 600 million euros ($645 million), and the company's valuation had soared to 5.8 billion euros ($6.2 billion). In 2025, according to a Reuters report on September 8th, in the latest Series C financing, Mistral AI successfully raised 1.7 billion euros ($2 billion). After the financing, its valuation is expected to reach as high as 12 billion euros ($14 billion), almost doubling compared to June last year. The leader of this financing is ASML, known as the "global lithography giant" and the hegemon in the semiconductor equipment manufacturing field. Its investment of 1.3 billion euros ($1.5 billion) has made it the largest shareholder of Mistral AI and obtained an important seat on the board of directors, once again making Mistral AI the focus of the global technology industry.
Besides ASML, Apple's favor for Mistral AI has long been evident. According to well - known Bloomberg journalist Mark Gurman, Apple is currently seriously considering acquiring French artificial intelligence startup Mistral AI, which may become Apple's largest acquisition in history. It's not hard to find that although Apple has been developing steadily in the AI field, in the face of increasingly fierce market competition, especially in the performance of intelligent assistants such as Siri, it is slightly inferior to its competitors. The strong technological potential shown by Mistral AI can exactly make up for Apple's shortcomings in AI, making the acquisition of Mistral AI an attractive option for Apple to enhance its own AI capabilities.
Technology Leadership, Multiple Models with Outstanding Strength
The reason why Mistral AI can obtain such a high valuation in a short time and attract the attention of many technology giants lies in its series of leading technologies and excellent model R & D capabilities. According to the company's official website, the Mistral series of models includes a total of 8 categories and is applied to 3 types of tasks:
* Support simple tasks that can be executed in batches, such as classification, customer support, or text generation, etc.
* Support intermediate tasks that require medium - level reasoning ability, such as data extraction, document summarization, email writing, job description writing, or product description writing, etc.
* Support complex tasks that require strong reasoning ability or high specialization, such as synthetic text generation, code generation, RAG, or Agent.
Mistral AI model series
Among its many series of products, Mistral AI has also achieved remarkable results in the exploration of lightweight and multimodal technologies.
Lightweight Models with Excellent Performance
Several models launched by Mistral AI have attracted wide attention in the industry. Among them, the open - source Mistral 7B model is truly amazing. Although this model has only 7 billion parameters, it shows performance that surpasses models of the same level. In many complex reasoning and coding tasks, the performance of Mistral 7B can even be comparable to some models with larger parameter scales. This lightweight design significantly reduces the demand for hardware resources during model operation and significantly improves the reasoning speed, enabling it to run efficiently on devices with limited computing power. This feature not only meets the current demand for optimizing device performance and energy consumption but also expands the possibilities for the application of AI technology in more scenarios. Paper address: https://hyper.ai/papers/2310.06825
On this basis, the improved sparse Mixture of Experts (MoE) model, Mixtral 8×7B, has become the most advanced MoE model on Hugging Face at that time. In most authoritative benchmark tests, Mixtral 8×7B has successfully surpassed well - known closed - source large models such as Llama 2 70B, and its reasoning speed has increased by more than 6 times compared with traditional models. This advantage enables it to quickly respond to user requests in practical applications, greatly improving the user experience. Under the Apache 2.0 open - source license, it has surpassed GPT - 3.5 Turbo, Claude - 2.1, Gemini Pro, and Llama 2 70B chat models in human benchmarks. Paper address: https://hyper.ai/papers/2401.04088
Exploration of Multimodal Technology
In addition to its in - depth development in the field of language models, Mistral AI is also actively deploying multimodal technology. Its Pixtral Large model has successfully achieved the integration of image understanding and text generation. In the medical field, this model can analyze and understand medical images and generate corresponding diagnostic reports or assist doctors in disease analysis; in the autonomous driving scenario, it can generate corresponding driving decision - making suggestions based on the road image information captured by the camera; in content creation, users only need to input an image, and Pixtral Large can generate relevant text descriptions based on the understanding of the image, providing new creative inspiration for creative workers. The development of multimodal technology has expanded the application boundaries of AI, enabling AI to play a role in more complex scenarios and also enabling Mistral AI to occupy an advantageous position in the diversified competition of AI technology. Paper address: https://hyper.ai/papers/2410.07073
Unique Open - Source and Rich Product Portfolio
In addition to its unique technological advantages, Mistral AI also attaches great importance to the construction of an open - source ecosystem and the layout of a product portfolio.
Open - Source Models Promote Community Collaboration
Several important models under Mistral AI, such as Mistral 7B and Mixtral, have their weights open and adopt the Apache 2.0 open - source license. Different from some AI companies that adhere to the closed - source route, Mistral AI firmly believes that open - source can gather the wisdom and strength of global developers. Through community collaboration, the model can be iterated and optimized quickly. In the open - source community, developers from all over the world can freely obtain the code and weights of these models, conduct secondary development, improvement, and application expansion according to their own needs and creativity, and create various AI applications suitable for different scenarios, including intelligent customer service, content creation tools, and educational auxiliary software.
Freely available weight models
The problems found and improvement suggestions put forward by developers during use, in turn, promote the continuous optimization and upgrading of the Mistral model, forming a virtuous cycle. This open - source model forms a sharp contrast with the closed - source routes of companies such as OpenAI and Anthropic. It not only contributes a large amount of valuable code and algorithms to the performance improvement and function enhancement of the model but also earns Mistral AI a good reputation and extensive community support.
Product Portfolio Covering Developers to Enterprises
Mistral AI's rich product portfolio provides a series of practical tools for developers and enterprises. A series of tools represented by La Plateforme and Codestral provide full - link support for developers from model customization to application development, reducing the threshold for developers to use AI technology. Developers can easily create intelligent agents with specific functions, perform personalized fine - tuning on models, and even get free basic completion capabilities, thereby improving the efficiency of code writing.
In terms of enterprise services, the enterprise intelligent assistant developed by Mistral AI can automatically handle customer consultations, answer customer questions quickly and accurately, greatly improving the efficiency of enterprise customer service and reducing labor costs. In terms of optimizing internal enterprise processes, this intelligent assistant can assist in document processing, improving office efficiency.
In addition, by launching the conversational AI assistant Le Chat for enterprises, enterprises can use Le Chat through subscription or authorization, connect to the enterprise's internal knowledge base, and provide customized solutions for customers. In just 100 days after its launch, Le Chat helped the company quadruple its business scale and achieved remarkable business results.
Core features of Mistral AI products
Analysis of the Reasons for Favoring Mistral AI
From technology to model, Mistral AI shows unique competitiveness, but this cannot fully explain why technology giants such as ASML and Apple are willing to invest or acquire it so generously. To uncover the underlying logic, it is necessary to further analyze it in combination with real - world needs and strategic considerations. After consulting multiple sources of information, the author makes the following speculations about the reasons.
Compensate for Apple's Shortcomings in AI Technology
Although Apple has been steadily developing in the AI field, its voice assistant Siri has gradually shown shortcomings such as insufficient natural language understanding ability and less - rich functions when compared with similar products of its competitors (such as Google Assistant and Amazon Alexa). The excellent performance of Mistral AI in language model technology, especially its advanced natural language processing ability (NLP) and efficient model architecture, can form a technological complement to Apple. By acquiring Mistral AI, Apple can quickly obtain top - notch AI technology and R & D teams, accelerate the iteration and upgrading of its own AI technology, improve the performance of products such as Siri, and enhance its competitiveness in the AI field.
In addition, its rich product portfolio and diverse technology application scenarios can also inject new vitality into Apple's product ecosystem. In terms of productivity tools, integrating Mistral AI's AI technology into Apple's office software (such as Pages, Numbers, Keynote) can achieve functions such as intelligent document editing and automatic data analysis, improving user office efficiency; in the enterprise service field, with the help of Mistral AI's enterprise intelligent assistant and customized solutions, Apple can further improve its enterprise - level service products (such as iCloud for Business, Apple Business Essentials).
Collaborate with ASML for Industrial Upgrading
As a leading enterprise in global semiconductor manufacturing equipment, ASML has always been committed to improving the performance and production efficiency of its lithography systems through technological innovation. In today's digital and intelligent era, the application potential of artificial intelligence technology in the semiconductor manufacturing field is huge. Whether it is the design optimization of lithography equipment or the parameter adjustment and monitoring during the production process, AI technology plays an important role. Through strategic cooperation with Mistral AI, ASML can deeply integrate its advanced AI models and algorithms and apply them to various aspects of its product R & D, production operation, and customer service. With the powerful data analysis and intelligent decision - making ability of AI technology, ASML is expected to further improve the performance of core equipment such as EUV lithography machines, achieve more precise lithography processes, shorten the equipment R & D cycle, and reduce production costs, thereby providing more advanced and efficient equipment and solutions for the development of the global semiconductor industry.
Conclusion
Although Mistral AI's market value and status are currently on the rise, it is not without controversy.
In August this year, it was reported that the company was involved in a plagiarism scandal. According to a former employee of Mistral AI, the company's latest model was suspected of being directly distilled from the DeepSeek model but was packaged as the result of self - reinforcement learning (RL) externally, and there were also cases of distorting benchmark test results.
In fact, as early as June 2025, AI expert Sam Paech found through technical analysis that the Mistral - small - 3.2 model was highly similar to DeepSeek - v3 in language pattern output, suggesting that the former may have "borrowed" the output style of the latter.
Susan Zhang, a researcher at DeepMind, even publicly criticized Mistral AI's unethical behavior.
However, among the many criticisms, there are also voices of approval. Clement Delangue, the co - founder and CEO of HuggingFace, once spoke up for it, asking, "Is it wrong to distill open - source models?"
