In a bid to carve out a niche in the increasingly competitive artificial intelligence (AI) landscape, France-based startup Mistral AI has unveiled its latest proprietary offering, Mistral Large. Positioned as a contender against major competitors, Mistral AI asserts that its new model boasts impressive capabilities that can rival leading players in the market.
According to a post shared by the Paris-based firm on February 26, Mistral Large demonstrated strong performance in a “multitask language understanding” test, outperforming several notable large language models (LLMs) except for GPT-4. Additionally, it showcased proficiency in various math and coding assessments.
Guillaume Lample, founder and chief scientist at Mistral AI, lauded Mistral Large as “vastly superior” to the company’s previous models. Notably, Mistral AI has introduced “Le Chat,” an AI chat interface built atop its models, akin to ChatGPT based on GPT-3.5 and GPT-4.
While Mistral Large’s prowess was not directly compared to newer entrants like xAI’s Grok and Google’s Gemini Ultra, released in November and February respectively, the firm remains confident in its product’s capabilities. Mistral Large, funded by a substantial $487 million investment from notable backers such as Nvidia, Salesforce, and Andreessen Horowitz in December, asserts its dominance in multilingual capabilities, with proficiency in English, French, Spanish, German, and Italian.
However, the decision to transition from open-source licensing for its earlier models to a closed, proprietary model akin to recent releases by OpenAI has sparked disappointment among some observers.
Despite the absence of assessments from third-party AI chatbot ranking platforms like Chatbot Arena, and Mistral Medium, the company’s previous model, ranks prominently among over 60 LLMs.
In a strategic move, Mistral AI has partnered with tech giant Microsoft to integrate Mistral Large into Azure AI Studio and Azure Machine Learning. This collaboration not only expands Mistral Large’s accessibility but also leverages Azure’s supercomputing infrastructure for training and scaling purposes. Eric Boyd, corporate vice president of Microsoft’s Azure AI Platform, emphasized the partnership’s focus on AI research and development.
Regarding pricing, Mistral Large offers competitive rates at $8 per million tokens of input and $24 per million tokens of output, positioning it marginally below the pricing of GPT-4 Turbo.
With Mistral AI’s valuation soaring to nearly $2 billion as of December, the firm’s foray into the AI market reflects a concerted effort to cement its position as a formidable player in the evolving landscape of artificial intelligence.
Image: Wallpapers.com
Disclosure Statement: Miami Crypto does not take any external funding, or support to bring crypto news to the readers. We do not have any conflicts of interest while writing news stories on Miami Crypto.