Reflection Debuts Beam AI Model, Mistral AI Launches Mistral Large 4
Reflection launched Beam, an open-weight AI model, while Mistral AI released Mistral Large 4.
"New AI models are dropping, promising big things for enterprises. Get ready for more customized, local AI solutions without breaking the bank."
Reflection has debuted Beam, an open-weight AI model designed to compete with Chinese models at a lower compute cost. The company aims Beam and future models at enterprises and sovereign nations, offering an "AI factories" product for customized, local AI systems by training Reflection's models on proprietary data. This move challenges China's open-weight AI lead.
Simultaneously, French AI lab Mistral AI has released Mistral Large 4, a new large multimodal model. Nicknamed 'Le Chonk' due to its one trillion parameters, Mistral Large 4 aims to rival both American and Chinese models. Mistral plans to make its weights available after safety testing, working with partners and governments to ensure responsible use. The model was trained using 4,000 NVIDIA GPUs, which is described as significantly less than some competitors.
Mistral hopes Mistral Large 4 will be a top open-weight model, particularly outside of China, and potentially outperform closed models in specific areas like cybersecurity, finance, and chip design. Reflection's focus on enterprise and sovereign nations with Beam suggests a strategic push for localized AI solutions, contrasting with Mistral's broader ambition to establish itself as a frontier AI lab.
These new open-weight AI models offer enterprises and sovereign nations alternatives for building customized AI systems with potentially lower compute costs. This development could impact strategic decisions regarding AI infrastructure and data sovereignty, providing more options for localized and secure AI deployments.
Find the right AI tool for your business
Chat with Insta and get matched to the right tool in seconds.
Try Insta Tool Finder →