See Usage Meter in Chat

Show the usage meter beside the chat box. The red bar would indicate overage

Show usage "cost" for the message in statistics

If I understand the change correctly, messages in long chats will “cost” more of the usage now. It would be great if there were an indictor what the last message actually costed.

Blacklist domains for search

I am not aware whether T3 Chat has some feature that prevents search from not reputable sources. Considering increasing amount of AI articles and such (example: Understanding T3 Chat Pricing: A Deep Dive Into Value and Features - Oreate AI Blog) it would be good if T3 Chat had some blacklisted domains (possibly disclosed) with option to request blacklisting or maybe even personal domain blacklist feature. Without relevant context even SOTA models can provide “text prediction slop” rather than useful results and this will future-proof T3 Chat.

Sort models by cost

Now that the model picker shows the relative model pricing, it would be a nice feature to be able to sort your favorite models by cost.

Rearrange favourited models

Rearrange favorite models and the overall model list.

Switch profile hotkey

I mean I could easily do this myself (with a simple chrome extension), but it would be nice to have a hotkey to switch between profiles.

Chrome extension

A chrome extension that you can have t3 chat in a side panel on any screen, maybe even have it be able to get tab context from like google docs easily

Model Request: glm4.7 flash

It has stable performance, please add it

Study mode (chatgpt)

The only thing keeping me from never directly paying OpenAI again is the need to use the ‘Study’ / ‘Study and learn’ mode during exam season. It is an amazing tool to help you learn concepcts and it would be amazing if it could be added to t3chat😀

Suggest updated models to replace old favourites

Currently, when new model versions release, users must manually unfavourite outdated models and favourite the new ones. This is repetitive and easy to forget. Proposed Solution: Add a subtle update indicator to the model selector that surfaces when newer versions are available. UI Placement: A button with an update icon positioned next to the filter button. Clicking it reveals a checklist of updatable models with checkboxes to select which ones to update in the list. State Visual New updates available Accent-coloured icon, subtle pulse or badge Viewed but not updated Muted text colour, static icon All models current Hidden entirely This keeps the interface clean while making the update path discoverable without being pushy.

Some latex rendering

I know this isn't probably common for normal people and LaTeX rendering is hard, but just leaving it here as a suggestion.

Search and use skills from skills.sh

Useful when Claude code is not needed like marketing, sales, social media, content writing, etc.

Shiny UI profile image

the profile image in the bottom left is the shiniest thing in the UI, it might be a little distracting.

Model Request: qwen 3.5

The recently released 3.5 is much smarter than the Qwen 3 235b, which is also less efficient. Why it hasn't been deprecreated yet is beyond me. Could you pls add Qwen 3.5 thinking/non-thinking?

Model Request: Mercury 2

https://www.inceptionlabs.ai/blog/introducing-mercury-2 https://x.com/StefanoErmon/status/2026340720064520670?s=20 Reason to add autoregressive LLM performance, but unbelievable speed

Allow BYOK models without key, now that usage is based on model price

Now that usage scales with model cost, could more expensive, currently key-locked models be allowed for everyone? They could have proportionally higher usage impact to account for their higher costs, while giving everyone the chance to use them. Could be combined with allowing keys for all models, as others have suggested, to give people the option to use all models within reason, or bring their own keys for more.

Restore the Delete button when hovering threads in list

The Delete button (now replaced with an Archive button) while hovering a thread in the thread list should be reactivated. Archiving sounds great and I’ll probably use it, but I’d use the Delete button way more often and I’d like it back. I’m fine with keeping the Archive button there too. I guess the reason you replaced it instead of just adding Archive is so that actions from that screen are undo-able, but I have delete-happy mentality formed pre-recycle-bin-OSes, consequences be damned! If you should choose against it, I’ll suffer using the hotkey instead - just don’t take that away from me.

Enable api key for image models

Can we enable the use of API keys for image models like OpenAI or Google? Those models are more expensive, and I don’t want their usage to reduce my available quota for other text models.

Model Picker - Option to hide whole company/lab

Problem: Users are likely to not have interest in models from all companies, so having all companies as the top of the hierarchy in the model picker can be clunky. Having the list be long enough to scroll and having the labs in a non-obvious order adds additional friction. Users may have other personal reasons but wanting to fully exclude a certain company/lab from their use (due to quality, company policies, etc.). Proposal: Provide users to hide a company/lab/family from the list of icons on the left side of the model picker. Provide a ‘show all’ to temporarily override the visibility setting. Also hide the hidden models from search results by default, with a ‘show all’ temporary override. Related: https://feedback.t3.chat/p/feature-to-hide-certain-models-from-all-models-grid

UI resets overnight, forcing me to refresh the page

When I close my laptop for the night and come back the following morning, and navigate to an open t3.chat tab, I find that it’s reset. I’m still logged in, but none of my archived chats are visible, it’s prompting me to “Make sure you agree to our Terms and our Privacy Policy”, and most importantly, my preferred default model is no longer selected. Instead, the t3.chat logged-out default is selected (currently that’s Kimi K2). I’m on Google Chrome 144, on macOS 26.2.