Sakana AIContinuous Thought Machines
ctm-imagenet
This Continuous Thought Machine uses internal neural dynamics to refine an ImageNet-1K classification decision. It is a visual research model, despite the repository's text-generation tag.
Research release
Classification develops over internal processing steps
Sakana reports adaptive computation and attention patterns that move across an image during iterative refinement. The release studies how a model reaches a prediction, rather than generating a natural-language description.ctm-imagenet official model card
Working with this release
Practical guidance. Measure classification accuracy and processing steps together. Do not interpret an attention visualization as proof of a human-like explanation.
Specifications & access
| Detail | Value and conditions |
|---|---|
| Status | Research release |
| Access | Downloadable |
| Model identifiers | SakanaAI/ctm-imagenet |
| Training task | ImageNet-1K classificationctm-imagenet official model card |
| Architecture | Continuous Thought Machinectm-imagenet official model card |
| Terms | Apache 2.0; experimental prototype, not intended for commercial or critical deployment.ctm-imagenet official model card |
Related models
- ctm-maze-large
Sakana AI ยท Continuous Thought Machines
An iterative neural model trained on 39 by 39 mazes to predict movement sequences.
Launch date not confirmedResearch releaseDownloadable