I recently tried T3 Chat by Theo from t3.gg.
I have been a long-time fan of Theo’s content. As someone who enjoys working with AI, I consider him a great source of insight. His programming experience and honest takes constantly push me to try new tools and approaches.
Theo previously worked at Twitch, and that background really shows. One thing he understands extremely well is UX. He noticed a clear gap in the market: many LLM chat tools still have a surprisingly bad chat experience. You can have great models, but if the interface is bad, the whole thing feels off.
That is where T3.chat comes in.
The idea is simple. Provide a great chat experience while giving access to almost all major models in one place.
And honestly, it works.
After watching one of Theo’s videos offering the first month for $1, I decided to try it. The regular price is $8. Interestingly, pricing around “8” is psychologically effective. We do not mentally round it to 10, but closer to 5. Small detail, but clever. (might be relevant to OpenAI's new price tiers btw)
The chat experience is really solid. Clean UI, fast responses, and a well-thought-out flow. I have to admit, this is where T3 Chat shines.
What I really enjoyed was experimenting with different models and branching conversations. For casual or conversational topics, the default Kimi K2 model was surprisingly fun to use. Overall, the devil is in the details, and the team around Theo delivered an amazing piece of software.
For deeper or more complex topics however, I still preferred Anthropic’s Sonnet or Opus via my Claude subscription.
I primarily use Claude, mostly because of memory and Claude Code. And in the end, that is exactly why I decided not to continue my T3 Chat subscription, at least for now.
There is apparentely also an T3 Chat app coming soon, which is a nice addition nonetheless.
However, memory matters. A lot.
I have gotten used to persistent memory across conversations, and it is something I cannot live without anymore when working with LLMs. A dedicated code interface would also be a very interesting feature, but memory is the real deal breaker for me.
Funnily enough, with recent news about OpenAI potentially introducing ads, memory becomes even more important. You need memory to personalize anything, including ads. But that is a topic for another day.
T3 Chat is a very well-designed product. The UX is excellent, and the flexibility of switching between models is genuinely powerful.
If memory is not a hard requirement for you, I would absolutely recommend trying it.
For me, I will keep an eye on it and revisit once memory becomes part of the experience.
Well done to Theo and the team.