AI / NEWS BRIEF
Mistral Large 4 is available through a preview API, with open weights promised by month-end. Its benchmark and efficiency claims still require independent testing.
Mistral has opened a public API preview of Mistral Large 4, a natively multimodal mixture-of-experts model with one trillion total parameters and 49 billion active parameters. The company says the model was trained from scratch on 3,800 Nvidia Grace Blackwell GPUs in its European data centres. It plans to release the weights by the end of October, alongside more architecture, benchmark and post-training details.
Developers can try the preview through Mistral Studio. The launch page lists API prices of $1.36 per million input tokens and $4.18 per million output tokens. Mistral says the model will ultimately support private-cloud and on-premise deployment. Until the weights arrive, however, the preview is not yet the downloadable open-weight release that the company has promised.
A capable European open-weight model would give organizations another route to self-hosting and regional control. Mistral reports strong results across coding, cybersecurity, agentic work and multimodal tasks, but most comparisons are presented by the provider and some technical details remain forthcoming. Independent evaluations should test the released weights, exact inference settings, operating cost and safety behaviour before the model is ranked conclusively or assigned any AIQuantumScore proprietary score.
Beam targets coding, reasoning and agentic work with 23 billion active parameters. Early access is limited, while weights and technical materials are promised later in October.
OpenAI will test visual ads during image generation and add third-party conversion and brand-suitability tooling, deepening ChatGPT’s advertising business.