French AI firm Mistral launches its first reasoning model to rival OpenAI and DeepSeek

Mistral, a Microsoft-backed French AI firm, released its first reasoning model on June 10, taking on China’s DeepSeek and OpenAI. Mistral’s CEO, Arthur Mensch, said the new reasoning model had the “specificity” of being able to reason in multiple languages.

Today, the CEO of Mitral, Arthur Mensch, said his company was unveiling a new reasoning model called Magistral, which was very competitive against all the other models like OpenAI’s o1 and DeepSeek’s R1. It also had the “specificity” of reasoning in multiple European languages. 

Mensch added that Mistral’s new model could execute more complicated tasks through a step-by-step logical thought process, and it was great at mathematics and coding. The Mistral team disclosed that their AI system specialized in open-weight large language models, which made their fundamental “parameters” publicly available. Developers can access and modify the model’s core knowledge and bypass the high costs and time of training a system from scratch.

Mistral launches seven successive AI models in 2025

Mistral CEO Arthur Mensch announced Magistral, their first-ever reasoning model, built to compete head-on with OpenAI’s o1 and DeepSeek’s R1.

But here’s the twist:

Magistral is optimized for European languages, making it a standout in a field dominated by English and… pic.twitter.com/rLIzuiLiqV

— Wes Roth (@WesRothMoney) June 10, 2025

Today’s launch of the Magistral reasoning model marked the seventh consecutive launch of an AI model upgrade from the French AI firm this year alone. The AI company recently unveiled Devstral, its agentic LLM for software engineering tasks, on May 21. Devstral was built under a partnership with All Hands AI, and it allegedly outperformed all open-source models on SWE-Bench Verified by a large margin. It was released under the Apache 2.0 license. 

On May 7, the AI firm pushed the boundaries of efficiency and usability of language models even further with the Mistral Medium 3 model. The model delivers high performance at 8X lower cost with simplified enterprise deployments.

Mistral released the Mistral Small 3.1 model on March 17, and it came with improved text performance, multimodal understanding, and an expanded context window of up to 128k tokens. The model, also released under an Apache 2.0 license, outperformed comparable models like Gemma 3 and GPT-4o Mini while delivering inference speeds of 150 tokens per second.

Mistral Saba debuted on February 17, and the 24B parameter model was custom-trained on curated datasets from across the Middle East and South Asia. The model provides more accurate and relevant responses than models over five times its size while being faster and cheaper. 

Mistral Small 3 dropped on January 30, and the latency-optimized 24B-parameter model was also released under the Apache 2.0 license. The model matches larger examples like Llama 3.3 70B or Qwen 32B and is an alternative for proprietary models like GPT4o-mini.

Codestral 25.01 landed on January 13, and the coding model is allegedly proficient in over 80 programming languages. The model is optimized for low-latency, high-frequency use cases and supports tasks such as fill-in-the-middle (FIM), code correction, and test generation.

Mistral contemplates a $1B funding round this year

The French AI firm confirmed that it was contemplating a $1 billion fundraising round in 2025. The forthcoming $1 billion fundraising round will equip the company with the necessary resources to implement its growth strategy, uphold its independence, and continue to challenge Silicon Valley’s supremacy in the global AI landscape. 

Initial talks with investors have already begun, with the official fundraising campaign expected to start later this year. The timing aligns with ongoing investor interest in the AI industry. Major investors include General Catalyst, Andreessen Horowitz, Nvidia, Salesforce, IBM, and Microsoft, with Microsoft also partnering to distribute Mistral’s models through Azure. The strong financial backing and strategic partnerships have attracted prominent European clients like BNP Paribas, AXA, Stellantis, and CMA CGM.

The $1 billion fundraising initiative will finance significant infrastructure developments, including a French data center, to strengthen data sovereignty and ensure regulatory compliance. The AI firm plans to expand globally with new offices in the U.S. and Singapore, targeting Asia-Pacific markets. CEO Mensch declared Mistral “not for sale,” instead preparing for a future IPO. 

The fundraising effort came as Mistral projected annual sales exceeding $100 million for the first time, marking its transformation from a research-focused startup to a commercially viable business. Mistral’s fundraising journey began when the company secured €105 million in seed funding a month after launch, followed by a €385 million Series A round in December 2023, reaching a $2 billion valuation. Its €600 million Series B in June 2024 pushed the valuation to nearly €6 billion, making it Europe’s highest-valued AI startup.

Cryptopolitan Academy: Tired of market swings? Learn how DeFi can help you build steady passive income. Register Now