Mistral AI challenges GPT-4 with "Large": An overview of the new model

Mistral AI, a Parisian AI startup, has entered the ring with its new offering, "Large," a robust neural network designed to compete with the likes of GPT-4 and Claude 2. Unlike its closed-off competitors, Large boasts minimal censorship and a usage-based pricing model, making it an attractive option for those seeking greater control and flexibility.

Large offers performance comparable to industry leaders like GPT-4, focusing on cost-effectiveness. Its pricing aligns closely with OpenAI 's $8 per million input tokens and $24 per million output tokens. This stands in contrast to GPT-4 Turbo's $10 and $30 for the same, potentially offering cost savings of approximately 25%.

Mistral AI 's Large also boasts a generous context window of 32,000 tokens, allowing it to consider a broader range of information while generating text, leading to potentially more nuanced and comprehensive responses. Additionally, it supports five languages - English, French, Spanish, German, and Italian - catering to a broader user base.

As a comparison, GPT-4 Turbo, which has a 128k-token context window, currently costs $10 per million input tokens and $30 per million output tokens. So Mistral Large is currently 1.25x cheaper than GPT-4 Turbo. Things are changing at a rapid pace, and AI companies update their pricing regularly.

While Mistral AI claims Large ranks second only to GPT-4 on various benchmarks, real-world performance may differ. Independent testing is crucial to determine how Large indeed compares to its competitors.

Math & Coding

Mistral Large demonstrates potential in coding and math-oriented tasks. See its performance on popular benchmarks below, measured against other top-performing large language models.

Le Chat: A New Conversational AI Option

Alongside its "Large" model, Mistral AI is introducing Le Chat, a chat assistant currently in beta testing.

Le Chat offers a free (for now) experience with the option to choose from three models: Mistral Small, Large, and Next (a concise prototype model). It's important to note that Le Chat currently cannot access the internet during interactions.

Looking ahead, Mistral AI plans to introduce a paid enterprise version of Le Chat with features like central billing and customizable moderation tools.

A partnership with 微软

Adding to its "Large" model and Le Chat announcement, Mistral AI strengthens its reach through a new partnership with Microsoft. This collaboration will provide access to Mistral models for Azure customers, adding to Microsoft's existing offerings.

While this integration seems like a straightforward addition to the Azure model catalog, it potentially signifies a deeper connection. The partnership opens the door for future collaboration opportunities and potential developments between both companies.

For Mistral AI, this alliance brings the benefit of expanding its customer base through the established distribution channel of Azure.

Microsoft, on the other hand, reinforces its commitment to an open partnership strategy on Microsoft Azure . While the company holds an investment in OpenAI's capped profit subsidiary, it also welcomes diverse AI models, such as Llama, onto its platform. This strategy aims to retain Azure customers within its ecosystem, potentially addressing any concerns regarding anti-competitiveness.