Mistral AI Launches Le Chat and Competes with the Best

Feb 28, 2024 | Chatbots, LLMs

Paris-based startup Mistral AI is progressively building an alternative to OpenAI, Google, and Anthropic, as demonstrated by its latest announcement. The company is launching a new flagship large language model called Mistral Large. In terms of reasoning capabilities, it is designed to compete with other leading models such as GPT-4, Gemini, and Claude 2.

In addition to Mistral Large, the startup is also launching its own alternative to ChatGPT with a new service called Le Chat. This chat assistant is currently available in beta version.

Mistral AI is known for raising funds in a very short time to develop foundational AI models. The company was officially created in May 2023. In December, the company closed a $415 million funding round, with Andreessen Horowitz (a16z) leading the round.

Founded by former DeepMind from Google and Meta employees, Mistral AI initially positioned itself as an AI company with a focus on open source. Although Mistral AI's first model was released under an open source license with access to model weights, this is not the case for its larger models.

Mistral AI's business model increasingly resembles that of OpenAI, as the company offers Mistral Large via a paid API with usage-based pricing. It currently costs $8 per million input tokens and $24 per million output tokens to query Mistral Large. In artificial language jargon, tokens represent small pieces of words – for example, the word "TechCrunch" would be divided into two tokens, "Tech" and "Crunch," when processed by an AI model.

By default, Mistral AI supports context windows of 32k tokens (typically more than 20,000 words in English). Mistral Large supports English, French, Spanish, German, and Italian.

In comparison, GPT-4 Turbo, which has a context window of 128k tokens, currently costs $10 per million input tokens and $30 per million output tokens. Thus, Mistral Large is currently 20% cheaper than GPT-4 Turbo. Things change rapidly and AI companies regularly update their pricing.

The "Mixture of Experts" Technology

The "mixture of experts" technology, offered by Mistral AI, also known as MOE, is an innovative approach to artificial intelligence. It involves dividing an AI model into multiple experts and activating only the most relevant experts based on demand. This saves resources and computing power while maintaining high performance. Mistral AI has made waves in the AI field by highlighting this revolutionary technology.

Positive Benchmark Results

Mistral AI's MISTRAL large model stands out for its cutting-edge reasoning capabilities and modularity. With a contextual window of 32,000 tokens, it can process vast amounts of information and produce relevant responses. Additionally, Mistral recently announced a partnership with Microsoft to make its large model available through Azure tools, strengthening its visibility and accessibility.

Comparing Mistral large's performance with other models such as GPT-4 and GPT 3.5, we find that Mistral offers relevant and diverse responses, particularly in areas such as website creation, analogical thinking, and programming. While some benchmarks highlight GPT-4's superiority in certain areas, Mistral large positions itself as a high-quality alternative, especially for businesses and AI professionals.

But how does Mistral Large compare to GPT-4, Gemini, and Claude 2? As always, it's very difficult to say. Mistral AI claims it ranks second after GPT-4 on several benchmarks. But there may be favorable benchmark selection and disparities in real-world usage.

An Alternative to ChatGPT

Mistral AI is also launching a chat assistant called Le Chat today. Anyone can sign up and try it at chat.mistral.ai. The company says it's a beta version for now and there may be "weird results."

Access to the service is free (for now) and users can choose between three different models – Mistral Small, Mistral Large, and a prototype model designed to be brief and concise called Mistral Next. It should also be noted that Le Chat cannot access the web when you use it.

Mistral also plans to launch a paid version of Le Chat for pro customers. In addition to centralized billing, companies will be able to define moderation mechanisms.

Partnership with Microsoft

Finally, Mistral AI announces a partnership with Microsoft. In addition to Mistral's own API platform, Microsoft will provide Mistral models to its Azure customers.

And yet, this also means that Mistral AI and Microsoft are now beginning discussions for collaboration opportunities and potentially more. The first advantage of this partnership is that Mistral AI will likely attract more customers with this new distribution channel.

As for Microsoft, the company is the main investor in OpenAI's capped-profit subsidiary. But it has also welcomed other AI models on its cloud computing platform. For example, Microsoft and Meta are partnering to offer Llama language models on Azure.

This open partnership strategy is a good way to keep Azure customers in its product ecosystem. It could also help with antitrust scrutiny.

Focus on International Expansion

Internationalization represents a crucial aspect for Mistral AI, highlighting its ambition to position itself as a major player on the global artificial intelligence scene. By developing language models capable of understanding and generating text in multiple languages, including English, French, Spanish, German, and Italian, Mistral AI aims to break linguistic and cultural barriers, making its technologies accessible to a global audience. This multilingual approach is not limited to expanding its potential market; it also reflects a deep understanding of local and regional nuances, essential for providing relevant responses and solutions to users worldwide. By emphasizing internationalization, Mistral AI is not only seeking to compete with giants like OpenAI and Anthropic, but also to establish new standards of inclusivity and accessibility in the field of generative AI.