Verified, then logged.
Sun 22 Nov 2026 · 00:23 KST / 15:23 UTC한국어 | ENSearch /

01 / AI · Analysis

Thinking Machines Lab releases Inkling 2 Mini weights

Inkling 2 Mini runs its middle layers three times for every output token it generates.

Thinking Machines Lab released the weights of Inkling 2 Mini, a 26.4 billion parameter model for writing code, under the Apache 2.0 licence, Neural Brew reported.

Neural Brew reported that the release includes the model’s weights, the values it learned during training. The model runs some of its layers more than once for each output token.

The repeat step takes place in the middle of the model, according to Neural Brew.

Middle layers repeat for each token

Neural Brew reported that the repeat step happens each time the model generates a token. A token is a small unit of model output.

According to the report, the model has 24 layers and runs the middle 12 three times for each token.

Model inputs and context limit

Neural Brew reported that the weight files total about 20 GB. The model accepts text and images as inputs.

Neural Brew reported that the model supports a context of up to 1 million tokens. The context limit sets how much information the model can handle at once.

AI-written by Doyun Lee