French AI company Mistral has released the weights of its one-trillion-parameter model, Mistral Large 4. Businesses and government agencies can download, customize and run it on their own infrastructure. The model is designed to strengthen cybersecurity by identifying software vulnerabilities, analysing threats and helping develop fixes.
Its open-weight approach gives organizations greater control over sensitive data and reduces their dependence on external cloud providers. Mistral also highlights improved coding performance and lower computing costs. However, the release also comes with risks. Once freely available, access can no longer be restricted, meaning malicious actors could also benefit from the technology’s capabilities.