Meta AI gets a Thinking mode for answers that require more steps

Meta has released Muse Spark 1.1, a new version of the company's AI model. It is now in a mode called Thinking in the Meta AI app and on meta.ai. Thinking here means that the model is allowed to work more gradually before it responds.

Meta describes the model as built for tasks where the AI needs to plan, use digital tools and understand both text and images. The company highlights coding and computing, among other things, but the comparisons with other models are based on Meta's own tests and are not independent proof.

For developers, Muse Spark 1.1 is also available through a public preview of the Meta Model API, an interface that allows other programs to send data to the model. A preview is a test version and can be changed before the service is completed.

Meta does not specify in the release exactly which countries, languages or subscriptions will get Thinking mode first. It is therefore not possible to promise that the function is already visible for every Swedish account. The price for API usage is also not clear from the launch article.

Därför spelar det roll

When an AI is given more time to plan, it can become more useful for multi-step tasks, but the answer still needs to be checked. For a small business, a safe test might be to let the AI suggest a work order for a project without giving it sensitive customer data or the right to publish anything.

Det här kan du göra

1. Choose a limited task with several steps, for example a content plan from your own, harmless base.
2. Ask the AI to show its work plan and what assumptions it makes before writing the final output.
3. Check the facts, figures and any proposed action before using or sharing the material.