Reflection released Beam, an open-weight model claiming lower cost and faster inference. Mistral launched Mistral Large 4 with 1 trillion parameters. US and European labs are catching up to Chinese open-weight models. Open-weight models offer enterprises more choices and potential cost savings. Sovereign AI is supported but not the primary driver for most enterprises. Model performance is less important than platform capabilities. Frontier AI model platforms provide orchestration, governance, and tooling. Open-source platforms are expected to emerge. Enterprises will likely use a mix of frontier and open-weight models. Platform capabilities should weigh as much as model choice.

• 1 trillion parameters for Mistral Large 4
• a fraction of the token cost and inference time for Beam