Mistral AI Unveils New Language Model and Announces Microsoft Collaboration
Summary:
France-based startup Mistral AI has unveiled its state-of-the-art large language model, Mistral Large. The firm claims the model outperforms prominent competitors in multiple tests, with exception of GPT-4. It also launched AI chat interface, "Le Chat". Funded by prominent names, Mistral's model, fluent in multiple languages, hasn't been compared yet with latest models like xAI's Grok and Google's Gemini Ultra. Despite earlier models being open-source, Mistral Large is a proprietary model, triggering some disappointment. Mistral recently partnered with Microsoft, enabling access to its studio through Azure AI Studio and Azure Machine Learning.
Mistral AI, a startup headquartered in France, has introduced its new cutting-edge large language model (LLM) to the already saturated AI market - asserting that its latest invention, Mistral Large, could outshine numerous major competitors. On February 26th, the company operating out of Paris disclosed that in a "multitask language understanding" evaluation, Mistral Large surpassed several prominent LLMs, with the only exception being GPT-4; it also demonstrated a promising performance in various math and coding examinations. However, no comparisons were made between Mistral Large and other models such as xAI's Grok or Google's Gemini Ultra, both of which were launched in November of last year and early February respectively.
Mistral's co-founder and chief scientist, Guillaume Lample, is confident that Mistral Large excels far beyond the previous models of Mistral AI. Alongside this, the company launched "Le Chat", which functions as an AI chat interface akin to how ChatGPT operates on GPT-3.5 and GPT-4.
Mistral, having secured $487 million in funding in December from established names like Nvidia, Salesforce, and Andreessen Horowitz, reported that Mistral Large has a vocabulary exceeding 20,000 English words and is multilingual, with fluency in French, Spanish, German, and Italian. Despite Mistral AI's initial model being made available under an open-source license, Mistral Large is a proprietary model kept under lock and key, similar to the latest LLMs from OpenAI. This has caused some disgruntlement among observers on platform X.
While the AI chatbot comparison platforms, such as Chatbot Arena, have not yet evaluated Mistral Large, its predecessor, Mistral Medium, currently ranks six out of more than 60 LLMs. Chatbot Arena determines these rankings through thousands of pairwise ratings, which are processed through a Bradley-Terry model that utilizes random sampling to arrive at an "Elo" score, predicting the winner in a head-to-head showdown between different models.
Mistral has also recently unveiled its collaboration with Microsoft, which will enable access to Mistral Large through Azure AI Studio and Azure Machine Learning. "Our collaboration with Microsoft solidifies our position in the market," Mistral AI stated regarding its commercial-focused LLM. Mistral Large will utilize Azure's "supercomputing infrastructure" for training and upsizing purposes. Further, both companies plan to collectively contribute to AI research and development as disclosed on February 26th by Eric Boyd, the corporate vice president of Microsoft’s Azure AI Platform.
As for its cost-effectiveness, Mistral Large is slightly less expensive than GPT-4 Turbo, with input and output rates of $8 and $24 per million tokens respectively, compared to GPT-4 Turbo's $10 and $30. As of last December, Mistral was valued nearly at $2 billion according to Bloomberg.
Published At
2/27/2024 7:40:49 AM
Disclaimer: Algoine does not endorse any content or product on this page. Readers should conduct their own research before taking any actions related to the asset, company, or any information in this article and assume full responsibility for their decisions. This article should not be considered as investment advice. Our news is prepared with AI support.
Do you suspect this content may be misleading, incomplete, or inappropriate in any way, requiring modification or removal?
We appreciate your report.