[GH-ISSUE #12539] Fine grained tags on ollama.com. #54834

Open
opened 2026-04-29 07:33:29 -05:00 by GiteaMirror · 1 comment
Owner

Originally created by @rick-github on GitHub (Oct 8, 2025).
Original GitHub issue: https://github.com/ollama/ollama/issues/12539

Originally assigned to: @hoyyeva on GitHub.

The tags (tools, thinking, etc) displayed in the models landing page covers all of the models in the family. However, not all models have those capabilities. It would reduce user confusion if the tag list was restricted to just the supported capabilities when viewing the model card for a specific model. For example, currently deepseek-r1:8b displays tools capability when the model doesn't support it:

Image

The only tags that should be displayed here are thinking and 8b.

Originally created by @rick-github on GitHub (Oct 8, 2025). Original GitHub issue: https://github.com/ollama/ollama/issues/12539 Originally assigned to: @hoyyeva on GitHub. The tags (`tools`, `thinking`, etc) displayed in the models landing page covers all of the models in the family. However, not all models have those capabilities. It would reduce user confusion if the tag list was restricted to just the supported capabilities when viewing the model card for a specific model. For example, currently deepseek-r1:8b displays `tools` capability when the model doesn't support it: <img width="972" height="224" alt="Image" src="https://github.com/user-attachments/assets/826cf4eb-a76a-4079-a27a-43c439654328" /> The only tags that should be displayed here are `thinking` and `8b`.
GiteaMirror added the feature requestollama.com labels 2026-04-29 07:33:49 -05:00
Author
Owner

@maternion commented on GitHub (Oct 22, 2025):

This is actually very important, thank you for the issue frob. Another example would be the Qwen3 models, the models prior 2507 update have hybrid thinking and non thinking capabilities if I remember correctly, but the new ones are divided into Instruct and Thinking models separately. So anyone trying /nothink would think that it's an issue but they just need to use the Instruct variant.

<!-- gh-comment-id:3430432397 --> @maternion commented on GitHub (Oct 22, 2025): This is actually very important, thank you for the issue frob. Another example would be the Qwen3 models, the models prior 2507 update have hybrid thinking and non thinking capabilities if I remember correctly, but the new ones are divided into Instruct and Thinking models separately. So anyone trying /nothink would think that it's an issue but they just need to use the Instruct variant.
Sign in to join this conversation.
1 Participants
Notifications
Due Date
No due date set.
Dependencies

No dependencies set.

Reference: github-starred/ollama#54834