French Artificial Intelligence startup Mistral unveils Mistral 3 open-source models

French Artificial Intelligence startup Mistral unveiled Mistral 3, a next-generation family of open-source models that includes small dense models 14B, 8B, and 3B and a larger sparse mixture-of-experts called Mistral Large 3. The company said the release represents its most capable model to date and noted Microsoft backing.

On Dec. 02, 2025 11:09 AM ET, Ravikash Bakolia reported that French Artificial Intelligence startup Mistral unveiled its next generation of models under the name Mistral 3. The announcement, posted by the company and noted in the article, positions Mistral 3 as an open-source family of models following the startup’s prior releases. The piece identifies the launch as the company presenting multiple configurations aimed at different deployment needs.

Mistral 3, according to the report, comprises three state-of-the-art small, dense models and a larger sparse model. The small dense models are listed as 14B, 8B, and 3B. In addition, the company introduced Mistral Large 3, described in the article as its most capable model to date and characterised as a sparse mixture-of-experts trained. The article links to the company statement for details and frames the release as the latest generation in Mistral’s model lineup.

The report also notes that Mistral is backed by Microsoft and that the company described the Mistral 3 family in a public announcement on its site. The coverage is concise and factual, focused on the model names, sizes, and the introduction of the sparse mixture-of-experts architecture for Mistral Large 3. The article includes an image credit to Alex Cristi /iStock via Getty Images and references the company post for the full release information.

65

Impact Score

Finance artificial intelligence compliance in European financial services

The article explains how financial firms can use artificial intelligence tools while meeting European, United Kingdom, Irish and United States regulatory expectations, focusing on risk, transparency and governance. It details the European Union artificial intelligence act, the role of cybersecurity, and the standards and practices that support compliant deployment across the financial sector.

Artificial intelligence becomes a lever for transformation in Africa

African researchers and institutions are positioning artificial intelligence as a tool to tackle structural challenges in health, education, agriculture and governance, while pushing for data sovereignty and local language inclusion. The continent faces hurdles around skills, infrastructure and control of data but is exploring frugal technological models tailored to its realities.

Microsoft unveils Maia 200 artificial intelligence inference accelerator

Microsoft has introduced Maia 200, a custom artificial intelligence inference accelerator built on a 3 nm process and designed to improve the economics of token generation for large models, including GPT-5.2. The chip targets higher performance per dollar for services like Microsoft Foundry and Microsoft 365 Copilot while supporting synthetic data pipelines for next generation models.

Samsung’s 2 nm node progress could revive foundry business and attract Qualcomm

Samsung Foundry’s 2 nm SF2 process is reportedly stabilizing at around 50% yields, positioning the Exynos 2600 as a key proof of concept and potentially helping the chip division return to profit. New demand from Tesla Artificial Intelligence chips and possible deals with Qualcomm and AMD are seen as central to the turnaround.

Contact Us

Got questions? Use the form to contact us.

Contact Form

Clicking next sends a verification code to your email. After verifying, you can enter your message.