update
May 12, 2026
By Teun
Ollama v0.23.4 adds image modalities for vision models
Ollama released v0.23.4 on GitHub on May 12, tagging commit 3af1a00. According to the release page, the update adds image modalities for vision models in the launch/opencode branch change set #15922.
Ollama released v0.23.4 on GitHub on May 12, with the tag pointing to commit 3af1a008e2dd4740874f07968b373db3bc723344. The release was marked with GitHub’s verified signature, according to the release page.
The update is tied to a single change listed as “launch/opencode: add image modalities for vision models (#15922),” tagged by GitHub user hoyyeva. Ollama did not include a longer changelog in the release post, so the public release notes point mainly to that feature and the associated source archive downloads.
⚡ New to this?
This is news because Ollama is a tool many people use to run AI models on their own machines or servers. A “vision model” is an AI model that can handle images, and “image modalities” means the model can accept image input, not just text. That can affect how people build assistants that read screenshots, documents, or other visual data.
🦞 OpenClaw angle
If you self-host Ollama in automation workflows, test this release against any pipeline that sends screenshots or document images to a model. Update your input validation so image payloads are explicitly handled, logged, and size-limited before they reach the model. If you wrap Ollama behind an agent, add a multimodal smoke test to confirm the model still returns the format your downstream code expects.
Image modalities are the input types a model can accept. In this case, the change suggests Ollama is extending support for vision models so they can work with images, not just text, according to the release title and commit message.
The release page shows the new version as v0.23.4 and provides source code archives in zip and tar.gz formats. No additional bug fixes, performance changes, or breaking changes are listed in the release post itself.
For users running Ollama locally, the practical implication is that this release centers on image handling for vision workflows. That matters for anyone using self-hosted models that need to process screenshots, documents, or other visual inputs alongside text.
GitHub also shows the release as part of the ollama/ollama repository and links the tag to the source tree at v0.23.4. The tag was published by hoyyeva at 19:51 on May 12, according to the GitHub page.
Because the release notes are sparse, the safest reading is simple: v0.23.4 is a feature-focused release that adds image modalities for vision models. Users who depend on Ollama for multimodal work will need to check the repository or the linked commit for implementation details.
