Mistral AI Became a Microsoft Partner With Its Almost ChatGPT-level LLM Model Available

Last Updated on February 27, 2024 9:27 am by Laszlo Szabo / NowadAIs | Published on February 27, 2024 by Juhasz “the Mage” Gabor

Mistral AI Became a Microsoft Partner With Its Almost ChatGPT-level LLM Model Available – Key Notes:

• Mistral AI, a well-funded AI startup, partners with Microsoft.
• Introduces three new large language models and a chatbot service.
• Secured significant funding, showcasing rapid growth and potential.
• Offers proprietary models capable of multiple languages and tasks.
• Performs competitively in benchmarks against leading models.

Introducing Mistral AI: A New Competitor in AI

Mistral AI, a well-funded Artificial Intelligence startup and now a partner of Microsoft, has introduced three new large language models (LLMs) and a chatbot service to compete with OpenAI’s ChatGPT.

The company has also announced a new alliance with Microsoft Corp., the top investor of OpenAI, which will provide Mistral’s engineers access to supercomputing infrastructure on Azure. Microsoft will also make the startup’s models available to its cloud platform customers.

Mistral, based in Paris, was launched in May and secured a $113 million seed round just four weeks later. In December, it raised an additional $415 million from a consortium led by Andreessen Horowitz and Lightspeed Venture Partners.

Prior to this, the company’s product portfolio only consisted of two open-source language models with 7 billion and 46.7 billion parameters, respectively.

Mistral’s Innovative Large Language Models

Now, Mistral has expanded its LLM lineup with three proprietary models, including the Mistral Large, which can generate text in English, French, Spanish, German, and Italian, as well as solve math problems and code software.

The model can take up to 32,000 tokens in a user prompt, each representing a few letters or numbers. Mistral claims that this model is the second most advanced on the market, just behind GPT-4.

In a benchmark test involving four LLM reasoning tasks, Mistral Large performed only slightly worse than OpenAI’s flagship model.

In a separate evaluation, it outperformed Llama 2 70B, an open-source alternative to GPT-4 released by Meta Platforms Inc. last year.

Mistral AI Capabilities and “Le Chat”

Developers can access Mistral Large through an API, which allows for the creation of custom moderation policies and integration with external applications.

Software teams can use external databases to provide answers to user inquiries. The API also has the option to package the model’s output into JSON files, which simplifies data transfer between applications and reduces the workload for integrating an AI system’s output.

Mistral Large is being launched alongside Mistral Small, a less advanced LLM with lower pricing and latency. Despite its scaled-back features, it still outperforms Mistral’s previous flagship model in various reasoning tasks.

The company plans to offer its latest AI models in multiple ways, including through an API and a new ChatGPT-like service called “Le Chat“. This service will also provide access to a prototype model called Mistral Next, which is designed to be brief and concise.

In addition, Mistral’s flagship LLMs will be available on Azure through its partnership with Microsoft, which also includes access to supercomputing infrastructure for training and running inference workloads.

The two companies will also explore the potential for developing customized versions of Mistral’s LLMs for specific use cases.

Definitions:

• Mistral AI: It’s a rapidly growing artificial intelligence startup, focusing on developing advanced large language models and AI services.
• Microsoft: A global technology company that develops, licenses, and supports a wide range of software products, services, and devices.
• Microsoft Azure: A cloud computing service created by Microsoft for building, testing, deploying, and managing applications and services through Microsoft-managed data centers.
• API (Application Programming Interface): A set of protocols, routines, and tools for building software and applications, allowing different software entities to communicate with each other.

Frequently Asked Questions:

1. What is Mistral AI?
   Mistral AI is an AI startup that specializes in creating advanced large language models and AI services, partnering with major companies like Microsoft.
2. How does Mistral AI compare to ChatGPT?
   Mistral AI introduces models that aim to compete directly with OpenAI’s ChatGPT by offering advanced language understanding and generation capabilities.
3. What unique features does Mistral AI offer?
   Mistral AI provides proprietary models capable of handling multiple languages, solving math problems, and coding software, with competitive performance benchmarks.
4. How can developers utilize Mistral AI?
   Developers can access Mistral AI’s models through an API for custom application development and integration, enhancing their software solutions.
5. What partnership has Mistral AI formed recently?
   Mistral AI has formed a strategic alliance with Microsoft, gaining access to Azure’s supercomputing infrastructure and making its models available to Azure customers.