The latest AI news and updates of Mistral AI — AI company building open-weight and commercial language models focused on efficiency and performance. Covering Mistral AI's latest company news and launches from the past 90 days.
Mistral AISep 8Mistral AI raised €3 billion in a Series D funding round at a post-money valuation exceeding €21 billion. Led by Samsung Electronics with co-leads EQT’s Scaleup Europe Fund and PSG Equity, the capital supports infrastructure expansion, commercial growth, and frontier research. The funding scales compute capacity for open-weight models and sovereign AI development across 20 countries.
Mistral AIAug 25Mistral AI and HUMAIN announced a strategic collaboration to develop localized frontier AI models for Saudi Arabia and the region. The partnership focuses on cybersecurity, voice, and Arabic-language performance, supported by a joint go-to-market strategy for regulated industries. This collaboration, valued in the hundreds of millions of Euros, provides sovereign AI infrastructure and customizable open-weight models.
Mistral AIAug 4Mistral AI released Shieldstral, a 3B open-weights multimodal safety classifier that runs on a single 16GB NVIDIA GPU. The model accepts plain-language moderation policies at inference time, unifying text and image safety evaluation without retraining. It returns a calibrated safety score for prompts, responses, and images, matching or outperforming guard models up to 7x its size.
Mistral AIJul 21Mistral AI is expanding its European compute capacity, with Microsoft committing to leverage this infrastructure to host Mistral’s frontier and efficient models. This expanded partnership brings Mistral’s models across Microsoft’s AI platform, offering enterprises and regulated industries deployment options ranging from cloud to fully disconnected environments.
Mistral AIJul 8Mistral AI launched Robostral Navigate, an 8B model that enables robots to navigate complex environments using only a single RGB camera. It achieves 76.6% success on the R2R-CE unseen benchmark, outperforming multi-sensor systems. The model generalizes across wheeled, legged, and flying robots, and uses prefix-caching to reduce training token requirements by 22x.
Mistral AIJun 24Mistral AI released OCR 4, a document understanding model that extracts text alongside bounding boxes, block classification, and per-region confidence scores. Supporting 170 languages, the model achieved a 72% average win rate in blind human evaluations against leading systems. It is available via API, Amazon SageMaker, Microsoft Foundry, or as a self-hosted container for private infrastructure.
Frequently asked questions