@mentions feature for custom GPTs not working with GPT-5.1 on Web and in Android app

GPT builders
,
1

Hi, I am having a strange issue with the @mentions feature and I would like to know if this is a known bug or if it is just me.

On the web, @mentions only work if I am using a legacy model. For example, if I select any legacy model such as GPT-4o or GPT-4.1, I can type @ in a chat and the menu pops up showing my GPTs and I can pick one as usual. But if I switch that exact same conversation to GPT-5.1 or Auto, typing @ does absolutely nothing. No menu, no list, no connectors, no UI reaction at all. The only variable that changes is the model.

On the Android app (from the Play Store), it is even worse. Typing @ does nothing in any chat, no matter which model I choose. I tried GPT-5.1, Auto, GPT-4o and other legacy models. In all cases, @ behaves like a normal character and never opens the mentions menu.

I also noticed something odd in older conversations where I had used multiple GPTs in the past. Before, there was a small GPT name badge above the assistant messages, showing which GPT was responding in that thread. Those labels are now gone in those old chats. The GPTs themselves still exist in my pinned list and I can start new chats with them, so they were not deleted. It just looks like the link between those conversations and the GPTs is no longer visible and @mentions also do not work there anymore.

Here is what I have already tried:

Different accounts on the same device, same behaviour.

Different browsers on the web client.

Incognito or private mode with no extensions.

Clearing cache and cookies for ChatGPT.

No VPN or proxy.

The pattern is always the same: on the web, @mentions only work if I use a legacy model, and stop working as soon as I switch to GPT-5.1 or Auto. On the Android app, @mentions never work in any model.

Is this a known issue with GPT-5.1 and or the mobile app, or is there anything else I should try on my side?

9 Likes
4

Yeah I have the exact same, I use GPT’s in the same chat to chain them together creating workflows. This has ruined it :frowning:

4 Likes
5

I opened a support ticket on this and they acknowledged the problem, but would not commit on a fix. Pretty disappointing….chaining GPTs is a key feature for many of my users. Support response:

Hi there,

Thank you for sharing your feedback with us. We really appreciate you taking the time! While we can’t guarantee that your suggestion will be implemented or provide updates on its status, we have shared it with the appropriate team for review. The @ feature has not been deprecated and won’t be deprecated our engineering team will work towards re-enabling it for 5.1 models.

If you need help with anything else, feel free to reach out anytime. We’re here to help.

Best,

Tomi

OpenAI Support

4 Likes
6

Yes, OpenAI messed up. Again. I’m amazed how new releases are handled at OpenAI, it seems as if the teams are working in silos where the left hand doesn’t know what the right hand is doing. Since this feature is central, hopefully they’ll address it soon, especially since Google is currently overtaking them in other areas.

3 Likes
7

A temporary solution: If you switch the model to 40 during an ongoing chat, you can use the “@” function again. However, this only works with the resources of the old model.

1 Like
8

Hi all, thanks for the reports. We’re really sorry for the trouble here. We can reproduce this on our side and engineering is aware that @mentions seems to work inconsistently.


It is not currently prioritized for an immediate fix, but the issue is being tracked and we’ll share updates if that changes. Thank you for your patience while we sort this out.

5 Likes
9

Also need this working. I might as well do my workflow using azure foundry if this feature doesn’t work. Gemini 3 is impressive and claude has been great for creative content. @mentions is the only reason other than “openai has my chat history and it’s a pain to export and make it useful” that I’m paying a sub for.

I am now going to look for ways to make it less painful and see if i can trim my sub count.

2 Likes
10

Facing the same issue. Chaining and tagging custom GPTs is the only reason I have been using ChatGPT, along with multiple others in my organization. Requesting OpenAI team to fix this feature as soon as possible.

3 Likes
11

Got the same issue, disrupts all workflows for my clients. Hope this gets prioritized soon

2 Likes
12

This is absolutely a critical part of my worlflow and breaking this essentially breaks chatGPT for me. Actively seeking out competitors.

1 Like
13

Hi everyone, this should be fixed now! Let me know if you're still unable to use @mentions in custom GPTs and I'd be happy to take it back to the team.

1 Like
15

Hello,

Tested right now and it still doesn’t work. No update available on the Play Store either. The web still doesn’t work for GPT-5.1, only for the legacy non-reasoning models.

1 Like
16

Ah, thank you for clarifying. Looks like we're actively rolling this (@mentions of custom GPTs in gpt-5.1) out, so your workspace may not have gotten the feature yet.

I'll post here when rollout is complete, though you may get it sooner!

4 Likes
17

what about mobile app? thanks!

2 Likes
18

Still unable to use @ mentions for GPTs or Projects while using 5.1.

1 Like
19

@OpenAI_Support In the webapp tagging GPTs in 5.1 is still unavailable for me.

1 Like
20

We're still gradually rolling this out, but we should hopefully see this fully supported by the end of this week!

2 Likes
21

I want to update everyone: I just tested the @mention feature for Custom GPTs in GPT-5.1 on the web and it’s working again, both in Auto, Instant, and Thinking. However, in the ChatGPT Android app it’s still disabled. I don’t know if this issue will come back in GPT-5.2, since it hasn’t been released for me yet, maybe because of my country. Is there any estimate for when this will be fixed in the Android app as well? @OpenAI_Support

2 Likes
22

With the GPT 5.2 rollout, tagging @ custom GPTs is working on some accounts and not working yet on others. Also, will the feature be available for Go Plan (in India)?

23

I just tested it, and it’s working normally on GPT-5.2, but only on the web. In the app, it’s still disabled, with no expected timeline for this feature to return.