Skip to content

Pick your style: Denkbaar

Book a conversation

Articles

Knowledge

What are open weights?

Open weights are the released numbers of a trained model, so you can run it yourself.

A developer examines a server beside an open and a closed box.
Released weights make it possible to run a model yourself. Illustration: Denkbaar.

Open weights are the numbers that give a trained model its knowledge. Once they are released, anyone can run the model themselves.

What you read about it

  • On 17 March 2024 xAI published the weights of Grok-1 under Apache 2.0, a licence that also allows business use. It was a raw base model, not trained to follow instructions: Grok-1 released.
  • Meta’s Muse Spark got closed weights in April 2026, unlike the open Llama line: Muse Spark.
  • Kimi K3.1 Pro first came out through an API on 23 October 2026. A release of the weights was announced, but our article does not confirm it: Kimi K3.1 Pro.
An open box with a module stands beside a folder and a server.
With open weights, the model weights have been released. Illustration: Denkbaar.

Why it matters to you

Open weights are not the same as usable. For an open model, look at three things: the licence, the size (does it fit affordable hardware?) and whether it is trained to follow instructions. Also check per version what was really released: a maker’s earlier choices are no promise for the next model.

Two people check a compute card and a document beside a server.
Check the licence, size and ability to follow instructions. Illustration: Denkbaar.

Written on 13 November 2026, based on the articles and sources below.

Sources

  1. Open Release of Grok-1 (xAI)
  2. Meta Muse Spark (first MSL model, closed weights)
  3. Kimi K3.1 Pro API (weights promised 2026-11-04)

Topics: open-weightsmodels