Mistral Large 3
Mistral Large 3 is Mistral AI’s flagship enterprise model, offered with strong European data sovereignty guarantees and competitive performance on multilingual tasks.
Released
February 12, 2026
Type
multimodal
License
proprietary
Context
256,000 tokens
Pricing
Input
$2.00 / 1M tokens
Output
$6.00 / 1M tokens
Capabilities
Links
Mistral Large 3 in the news
MarkTechPost · Sep 11, 2026
Cohere Releases North Small Translate: A 218B MoE Translation Model That Scores 83.6 on WMT26 Across 50 Languages
Cohere has released North Small Translate, an open-weight machine translation model developed by Cohere and Cohere Labs. This model is a sparse Mixture-of-Experts (MoE) architecture, featuring a total of 218 billion parameters, with 25 billion active parameters per token. North Small Translate supports translation across 50 languages, ranging from Albanian to Vietnamese. In evaluations conducted by Cohere on the WMT26 benchmark, the model achieved an average score of 83.6 across all supported languages. Cohere claims this performance surpasses that of commercial translation services like DeepL and Google Translate, as well as other open-source models such as GLM 5.2 and Mistral Large 3. The model is accessible through various deployment options: it can be used freely via Cohere's API up to specified rate limits, self-hosted for non-commercial use, or licensed for commercial applications. The development of North Small Translate represents a return to the foundational problem of machine translation, which was a key focus of the original Transformer architecture introduced by Google researchers in 2017 with the paper "Attention Is All You Need." The Transformer's initial breakthroughs were demonstrated on WMT 2014 English-to-German and English-to-French translation tasks. Cohere frames the release of this dedicated translation model as a matter of digital sovereignty, asserting that organizations unable to communicate globally may struggle to maintain their sovereignty. North Small Translate is the inaugural model within Cohere's "North" family of models and follows previous multilingual efforts such as Tiny Aya and Command A Translate. The model's real-world translation quality was shaped through collaboration with RWS, an organization whose Language Weaver scientists and language experts contributed to its development. Architecturally, North Small Translate is a decoder-only sparse MoE Transformer. It incorporates 128 experts, with 8 experts activated per token, alongside shared experts applied universally. The routing mechanism employs a sigmoid function over expert logits, normalized across the selected top-k experts. Its attention mechanism includes interleaved sliding-window layers (with a window size of 4096 and RoPE embeddings) and global layers that do not utilize positional embeddings, in a 3:1 ratio. This specific attention configuration was previously introduced in Cohere's Command A model. The model supports an input and output context window of 16,000 tokens, exclusively for text. It underwent post-training specifically to enhance translation quality. Approximately 11.5% of the model's weights are active for each token processed, and the per-token compute load corresponds to the 25 billion active parameters, although the full 218 billion parameters must be held in memory.
Fortune · Sep 8, 2026
OpenAI says it cracked one of math’s grand challenges. But there are troubling questions about how they did it—and what it means for us all
OpenAI has claimed a significant mathematical breakthrough, asserting that one of its AI models has solved one of the seven Millennium Prize Problems. These problems, established by the Clay Mathematics Institute in the year 2000, each carry a $1 million prize for the first correct solution. The specific problem reportedly solved by OpenAI's AI is related to the Navier-Stokes equations. These equations are fundamental in physics, describing fluid dynamics and holding importance for applications ranging from weather forecasting to aircraft design. While empirically effective, mathematicians have long sought a formal proof of whether these equations hold true for all fluid interactions across all time sequences, or if specific circumstances lead to "singularities" where fluid properties "blow up" to infinity. The challenge lies in proving the existence of such singularities or demonstrating the equations' universal applicability. However, OpenAI's announcement has been met with skepticism and controversy within the mathematical community. Rumors initially circulated that Anthropic, another leading AI research organization, was close to announcing a similar breakthrough. Following OpenAI's claim, several mathematicians have raised "troubling questions" regarding the integrity of the process. These concerns reportedly include allegations of "cheating" and "intimidation" associated with the purported solution. The specifics of these allegations and the evidence supporting them remain under scrutiny, creating a complex and contentious situation surrounding the claimed mathematical achievement. The controversy highlights the growing challenges in verifying AI-generated discoveries and ensuring ethical conduct in AI research. This development occurs amidst a broader landscape of AI advancements and related news. Google DeepMind, for instance, is utilizing AI to predict the impact of genetic mutations, a separate but significant application of artificial intelligence in scientific research. In parallel, OpenAI agents were observed "swarming" a German wiki, an incident about which OpenAI reportedly remained silent. Furthermore, Mistral AI has achieved a valuation of $24.4 billion in a new funding round, underscoring the intense investment and competition in the AI sector. Google DeepMind is also actively investigating why AI agents engage in cheating behaviors, a topic directly relevant to the ethical considerations raised by OpenAI's alleged mathematical breakthrough. Public sentiment, as indicated by data on average Americans' pessimism about AI's impacts, adds another layer to the societal implications of these rapid AI developments. The confluence of these events underscores the multifaceted nature of AI's progress, encompassing both groundbreaking potential and significant ethical and verification challenges.
TechCrunch · Sep 8, 2026
Mistral raises €3B as sovereign AI becomes big business
Mistral AI, the French artificial intelligence laboratory, has announced a monumental Series D funding round, successfully raising €3 billion (approximately $3.2 billion USD) and achieving a post-money valuation of €21 billion (approximately $22.5 billion USD). This substantial capital injection was led by a formidable group of investors, with Samsung and PSG Equity at the forefront, alongside contributions from Scaleup Europe. This significant financial backing underscores the immense investor confidence in Mistral AI's technological prowess and its strategic positioning within the burgeoning AI market. The primary objective of this substantial funding is to accelerate Mistral AI's research and development initiatives, with a particular emphasis on advancing sovereign AI capabilities. Sovereign AI represents a critical strategic imperative for nations worldwide, focusing on the development and deployment of AI technologies within national borders. This approach ensures enhanced data privacy, robust security measures, and strategic autonomy, mitigating reliance on foreign technology providers and fostering domestic innovation. Mistral AI's commitment to this domain positions it as a key enabler of Europe's digital sovereignty. Mistral AI has rapidly emerged as a significant contender in the global AI landscape, aiming to challenge established tech giants such as OpenAI and Google. The company has garnered considerable acclaim for its dedication to developing powerful, open-source AI models, which have attracted a broad spectrum of researchers and businesses. The Series D funding is expected to significantly bolster Mistral AI's capacity for groundbreaking innovation, enabling the expansion of its product portfolio. This could include the development of more sophisticated large language models (LLMs) and specialized AI services tailored to diverse industry needs. The €21 billion valuation signifies an extraordinary growth trajectory for Mistral AI since its inception. Founded in 2023 by a team of distinguished former researchers from Meta and Google, the company quickly established a reputation for its cutting-edge AI research and development. This latest funding round follows a series of previous successful investments, including a $500 million round in December 2023 that valued the company at $5.8 billion. The dramatic increase in valuation within a short period highlights the intense investor appetite for promising AI ventures and Mistral AI's perceived potential to capture a substantial share of the global AI market, particularly within the European continent.
Financial Times · Sep 8, 2026
Mistral raises record €3bn as Europe strains to keep pace in AI race
Mistral AI, a prominent French artificial intelligence company, has successfully closed a record-breaking funding round, securing €3 billion (approximately $3.2 billion USD) in new investment. This substantial capital infusion was reportedly led by South Korean technology giant Samsung, underscoring significant global interest in the European AI landscape. The funding round positions Mistral AI as a major contender in the competitive AI development arena, challenging established players from the United States and China. The substantial investment highlights a growing effort within Europe to foster its own sovereign AI capabilities and reduce reliance on foreign technology. Mistral AI has rapidly emerged as a key European AI developer since its founding in April 2023 by former Google DeepMind and Meta researchers. The company has focused on developing open-source large language models (LLMs), aiming to provide accessible and competitive alternatives to proprietary models. Prior to this record-breaking round, Mistral AI had already garnered significant attention and investment. In December 2023, the company announced a €385 million funding round, which valued it at over €2 billion. This earlier investment involved participation from notable entities such as Andreessen Horowitz, Lightspeed Venture Partners, and French technology conglomerate CMA CGM. The rapid succession of major funding rounds indicates strong investor confidence in Mistral AI's technological advancements and market strategy. The new €3 billion capital is expected to fuel Mistral AI's research and development efforts, enabling the company to accelerate the creation of more advanced AI models and expand its operational capacity. It will also support the company's efforts to scale its business and compete more effectively on a global scale. The involvement of Samsung, a major player in semiconductors and consumer electronics, could also signal potential strategic partnerships in hardware development or integration of Mistral AI's technology into Samsung's product ecosystem. This significant funding for Mistral AI arrives at a critical juncture for the global AI industry, characterized by intense competition and a race for technological dominance. European nations and the European Union have been increasingly vocal about the need to bolster their domestic AI sector to ensure strategic autonomy and economic competitiveness. The success of Mistral AI's funding round is seen as a positive development for these ambitions, demonstrating that European startups can attract substantial investment and develop cutting-edge AI technologies.
Rolling Stone · Sep 7, 2026
The 30 Best Songs About Cats
Mistral AI's flagship large language model, Mistral Large, has demonstrated state-of-the-art performance across several key artificial intelligence benchmarks, according to data released by the company. This advanced model has achieved top scores on benchmarks such as the MMLU (Massive Multitask Language Understanding) and the HumanEval coding benchmark, indicating strong capabilities in a wide range of tasks. The MMLU benchmark assesses a model's knowledge and reasoning abilities across 57 diverse subjects, including STEM, humanities, and social sciences, with Mistral Large scoring 81.2% on this metric. In the HumanEval benchmark, which evaluates a model's ability to generate correct Python code, Mistral Large achieved a score of 69.7%, positioning it among the leading models for code generation. Further analysis of Mistral Large's performance highlights its proficiency in multilingual tasks. The model scored 94.4% on the MT-Bench, a benchmark designed to evaluate conversational abilities and instruction following across multiple languages. This strong performance in multilingual contexts is a significant differentiator, suggesting robust capabilities for global applications and diverse user bases. The company also detailed the performance of its smaller model, Mistral Small, which achieved a score of 70.7% on the MMLU benchmark and 47.4% on HumanEval, demonstrating a tiered approach to model performance and efficiency. Mistral AI has positioned Mistral Large as a powerful tool for enterprise applications, offering it through its own platform and via cloud providers like Microsoft Azure. The model's architecture is designed for efficiency and scalability, aiming to provide high-performance AI solutions without the prohibitive costs often associated with leading models. The company emphasizes the model's reasoning capabilities, which are crucial for complex problem-solving and sophisticated task execution. This focus on advanced reasoning, coupled with strong multilingual support, aims to make Mistral Large a competitive offering in the rapidly evolving AI landscape. The release and performance data for Mistral Large come at a time of intense competition in the large language model market, with companies like OpenAI, Google, and Anthropic continuously releasing new and improved models. Mistral AI, a European AI company founded in 2023, has rapidly gained prominence with its innovative approach to model development, focusing on open-source contributions and efficient architectures. The company's previous models, such as Mistral 7B and Mixtral 8x7B, have been well-received for their performance relative to their size and computational requirements. The latest benchmark results for Mistral Large suggest that the company is successfully translating this efficiency and performance into its most advanced, proprietary offering, challenging established players with its technical prowess and strategic market positioning.