Mistral AI and NVIDIA have jointly released a new language model for enterprises called Mistral NeMo 12B. This model demonstrates outstanding performance in areas such as multi-turn dialogue, mathematics, common sense reasoning, world knowledge, and programming. Mistral AI boasts that this model possesses unprecedented accuracy and flexibility in various enterprise applications. Mistral NeMo has a context length of 128K, enabling it to handle complex and lengthy inputs more effectively than many competitors. The model can process a large amount of information coherently, generating contextually relevant outputs for diverse enterprise needs. Guillaume Lample, co-founder and chief scientist of Mistral AI, emphasized the benefits brought by this collaboration: "By leveraging NVIDIA's hardware and software, we have developed a model that is highly accurate, flexible, and efficient, while also benefiting from enterprise-level support and security." The training of this model utilized NVIDIA's AI infrastructure, including 3,072 H100 80GB Tensor Core GPUs on the DGX Cloud AI platform. This process incorporated accelerated training techniques to optimize performance. Mistral NeMo offers several key advantages for enterprise users: - Versatility: Released under the Apache 2.0 license, it is designed as a plug-and-play alternative for systems using Mistral 7B. - Efficiency: Adopts the FP8 data format during inference, reducing memory requirements and accelerating deployment speed without sacrificing accuracy. - Multilingual capability: Demonstrates excellent performance in languages such as English, French, German, Spanish, Italian, Portuguese, Chinese, Japanese, Korean, Arabic, and Hindi. - Improved tokenization: Introduces the new Tiktoken-based tokenizer, Tekken, which shows efficiency improvements in over 100 languages. - Easy deployment: Packaged as the NVIDIA NIM Inference Microservice, allowing for quick setup in various environments. - Hardware flexibility: Can run on a single NVIDIA L40S, GeForce RTX 4090, or RTX 4500 GPU, striking a balance between performance and cost. For enterprises seeking to implement advanced AI capabilities, Mistral NeMo 12B offers a powerful combination of features. Its multilingual capability and efficient processing make it suitable for a wide range of enterprise applications. Furthermore, Mistral NeMo is designed with enterprise-grade security and support, including dedicated feature branches, rigorous validation processes, and comprehensive service level agreements. Enterprises can seamlessly integrate Mistral NeMo into their commercial applications and benefit from direct access to NVIDIA AI experts as well as reliable and consistent performance. Now, users can obtain Mistral NeMo, and a downloadable NIM version will also be released soon.