Mistral AI releases Mistral Large, a new LLM that approaches GPT-4 performance,Le Chat, a chatbot, demonstrates the capabilities of Mistral AI's models,Microsoft partners with Mistral AI to make Mistral Large available on Azure
Mistral AI Releases Mistral Large
French startup Mistral AI has released a new large language model (LLM) called Mistral Large that approaches the performance of GPT-4. The release of Mistral Large comes just six months after the company was founded and achieved a $2 billion valuation. Mistral AI's previous open-source LLM, Mixtral 8x7B, beat Meta's Llama 2 34B, causing excitement in the industry.
Mistral Large's benchmark figures show that it outperforms Claude 2, Gemini Pro, and Llama 2-70B. While the number of parameters for Mistral Large has not been disclosed, its 32k context window is equivalent to around 20,000 words. Mistral Large is available through API access only and is 20% cheaper than GPT-4, with input tokens costing $8 per million and output tokens costing $24 per million. This competitive pricing strategy could influence the broader market, especially as companies like IBM see significant gains in the AI rally, as discussed in "IBM shares surge 45%, outpacing Nvidia on AI rally".
Le Chat: Mistral AI's Chatbot
Mistral AI has also released a beta version of its chatbot, Le Chat. The chatbot allows users to experience responses from Mistral Large and offers access to Mistral AI's smaller models, Mistral Small and Mistral Next. Mistral Small outperforms Mixtral 8x7B and has lower latency, while Mistral Next offers more concise and cheaper interactions. The development of sophisticated chatbots like Le Chat highlights a broader trend, where AI agents are becoming increasingly capable, raising questions like "Will AI Agents Steal Your Job Or Help You Do It Better?"](/life/ai-agents-and-jobs).
Mistral AI has tried to make its models as useful and unbiased as possible. The company's tunable system-level moderation mechanism warns users in a non-invasive way when they are pushing the conversation in directions where the assistant may produce sensitive or controversial content. This approach aligns with the growing emphasis on responsible AI development, a topic extensively debated in regions with diverse regulatory landscapes, as explored in "North Asia: Diverse Models of Structured Governance".
Microsoft Partners with Mistral AI
Microsoft has entered into a multi-year partnership with Mistral AI and will make its Azure platform available for Mistral AI to develop and deploy its models. Arthur Mensch, CEO of Mistral AI, said, "We are thrilled to embark on this partnership with Microsoft. With Azure's cutting-edge AI infrastructure, we are reaching a new milestone in our expansion propelling our innovative research and practical applications to new customers everywhere." This partnership underscores the increasing collaborations between tech giants and innovative AI startups, a trend that is also seeing Microsoft team up with other nations for AI growth, as seen in "Singapore, Microsoft team up for AI growth".
What Does It All Mean?
Mistral AI's release of Mistral Large and Le Chat demonstrates the company's commitment to advancing artificial intelligence and AGI. With Microsoft's partnership and the availability of Mistral Large on Azure, Mistral AI is poised to make a significant impact in the industry. The release of GPT-5 may be imminent, but Mistral AI's multimodal functionality and competitive pricing make it a formidable competitor. The broader implications of such advancements are often discussed in the context of global AI development, including in reports like the Stanford AI Index Report 2024 AI Index Report.
Will Mistral AI's multimodal functionality and competitive pricing make it a bigger threat to OpenAI than the release of GPT-5?
You may like:
Opinion: Asia’s AI Revolution – A Journey of Growth, Challenges, and Promise,Mastering Your LinkedIn Profile with ChatGPT,7 Types of Artificial Intelligence Explained,Pivot into an AI job within 6 months!,Or read more on Microsoft's blog post announcing its partnership with Mistral AI.






Latest Comments (3)
given the 32k context window and its performance against models like Claude 2, I'm curious about the specific pre-training architecture and dataset mix. are they leaning heavily on a particular type of data, or is it a more general web-scale corpus similar to what we see with other large LLMs?
The 20% cost reduction for Mistral Large against GPT-4 is significant. For financial institutions in Hong Kong, where data volume is immense and regulatory compliance adds layers of checks, even a slight edge in price performance can impact large-scale LLM deployments. The market will certainly shift if those figures hold up.
this is so smart how they're pricing Mistral Large, 20% cheaper than GPT-4, especially the output tokens. in luxury, every Euro counts on the margins, and for creative agencies making a ton of variations, those output costs add up fast. very strategic for the european market.
Leave a Comment