A PERSONAL JOURNAL

Thoughts on AI from the brighter side

Positive, clear and useful stories about AI — for everyday life, curiosity and the future.

Posts tagged #öppna-modeller 06

Two connected computers share a large model, represented by floating blocks. AI-generated illustration.

In brief16 October 2026

Muse Spark Open is downloadable – but needs serious memory

Meta has released a large AI model with downloadable weights. Apple's example shows both the possibility of running it locally and how far it is from an ordinary home computer.

On October 15, Meta released Muse Spark Open. Unlike Muse Wonder, which we recently covered, its model weights are downloadable. Weights are the learned numbers needed to run a model yourself. This is not a small chat app, though. The model is mainly designed for coding and multi-step tasks. Apple…

Read the post ↗

A smaller AI module stands in front of a half-open door while larger blocks wait behind it. AI-generated illustration.

In brief7 October 2026

Large 4 is arriving in two stages

Mistral has opened a paid preview of its largest model so far. Anyone who wants to download the model weights and run it themselves will have to wait until later in October.

Read the post ↗

In brief6 October 2026

Beam is unveiled – but not released

Reflection has now introduced its open-weight AI model Beam. But access is the important detail: for now, only a select group can try it.

Read the post ↗

In brief4 October 2026

Open weights are not the whole recipe

Reflection is preparing a new AI model with open weights. The news is also a useful reminder: open weights and open source are not quite the same thing.

Read the post ↗

In brief20 September 2026

Gemma and Gemini – similar names, different things

Google’s AI names are easy to mix up. One simple distinction helps: Gemini is available as a service, among other things, while Gemma offers open models to build with.

Read the post ↗

In brief12 September 2026

What does an open AI model mean?

Open models give more people a chance to experiment. But open can mean different things, and conditions still apply even when a model is available to download.

Read the post ↗