The French startup Mistral has launched a new AI model, Mistral Large, positioning itself as a significant competitor to OpenAI's GPT-4, with an MMLU score of 81% compared to GPT-4's 86%. The model, described as occasionally outperforming GPT-4 in terms of logic and having "light" censorship, is now available on platforms such as OpenRouter and Lmsys arena. Despite being open-source, critics argue that Mistral's reliance on closed-source breakthroughs undermines its innovative claims. Comparisons with other models like Claude 2 and Gemini 1.5 Pro, which is priced 20 times cheaper than GPT-4, highlight the competitive landscape of AI development. Mistral's launch has sparked discussions about its performance, pricing, and the broader implications for the AI industry, contributing to Paris's reputation as a vibrant AI tech hub.