Mistral AI, a Paris-based startup, is shaking up the AI industry with its Mixtral 8x7B model, outperforming giants like ChatGPT and Llama with speed and efficiency.
Mistral AI steps into the competitive field with Mixtral 8x7B, an open-source model leveraging Sparse Mixture of Experts architecture for efficient pre-training and scalability. The model and its “Mixture of Experts” approach achieves the performance of larger models with far less compute and far less time required for training.
Demonstrating impressive performance, Mixtral 8x7B matched ChatGPT 3.5 in testing and outdid Llama 2 70B on benchmarks, with six times faster inference.
With significant funding secured, Mistral AI introduces ‘La Plateforme’ for model API endpoints, offering Mistral Tiny and Small, and monetizing the forthcoming Mistral Medium. This dual-revenue strategy boosts Mistral’s open-source venture, propelling its valuation potentially over two billion dollars.
Sources include: Analytics India