With the release of the new Mixtral 8x7B model, Mistral AI has opened up a beta version of "La plateforme" to developers, providing convenient access to their cutting-edge generation and embedding models. The initial service showcases the efficient deployment and customization capabilities of using Mistral's open models. The launch of this platform includes three main services: 1. Mistral-tiny: The most cost-effective endpoint, Mistral-tiny runs on upgraded base models with a higher context length (now 32k) and achieves a score of 7.6 on MT-Bench. Currently, it only supports English and is designed for cost-sensitive applications. Mistral 7B Instruct v0.2. 2. Mistral-small: This endpoint utilizes the newly released models and excels in code generation, mastering five languages - English, French, Italian, German, and Spanish. With a score of 8.3 on MT-Bench, Mistral-small offers a balance between performance and affordability. Mistral 8x7B Instruct v0.1. 3. Mistral-medium: Representing the pinnacle of quality in Mistral AI's product line, Mistral-medium uses a high-performance prototype model that surpasses GPT-3.5 in all metrics. It is on par with Mistral-small in language and code capabilities but surpasses it with a higher MT-Bench score of 8.6, making it an ideal choice for applications that demand the highest quality. In addition, Mistral AI introduces Mistral-embed, an embedding model with 1024-dimensional embedding vectors designed for retrieval applications. It demonstrates a powerful retrieval score of 55.26 on MTEB, catering to a wide range of data retrieval and analysis needs. The Mistral platform aims to empower developers with cutting-edge generative AI. The API follows the popularized chat interface standards used by other services to minimize migration barriers. Python and JavaScript client libraries are provided. As Mistral transitions from closed testing to general availability, they plan to continue expanding capacity and services. Interested users can now register for early access. Mistral's ultimate vision is to enable the community to access original models that will drive the next wave of AI innovation. Launching powerful embedding and generation endpoints is the first step towards realizing this roadmap.