A hosted multi-model AI chat app created by YouTuber Theo Browne and co-founder Mark. It puts frontier models from OpenAI, Anthropic, Google, and others into one fast, well-designed interface with search, attachments, profiles, and shareable new-chat URLs with query parameters for model and search state. It is explicitly a chat product, not an agent orchestration workspace—T3’s separate coding product is T3 Code. Pricing is usage-metered with a Base bucket (refills every 4 hours) and an Overage bucket (refills monthly); a Pro tier exists around $8/month, and a Premier tier at $50/month adds roughly 10x usage. Users can also bring their own OpenRouter key to pay per request directly. T3 states it does not train its own models and opts out of provider training where possible. Trustpilot shows 3.5/5 from a small base (4 reviews in the last 12 months), with complaints concentrated on the 2026 usage-limit changes and a billing/cancellation issue, alongside praise for speed and UX.
Pros & Cons:
Pros: One hosted interface for switching between frontier models quickly, which is its core value over provider-native apps; usage model uses a Base bucket that refills every 4 hours, so short sessions feel less “rationed” than a monthly message counter; BYOK via OpenRouter gives cost-control and transparency for power users; states it does not train on user data and opts out with providers where possible; clean, fast UX is the most consistent praise across reviews; shareable /new URLs with model and search query parameters are useful for reproducibility.
Cons: Usage-limit changes in 2026 drew sharp criticism—a Trustpilot reviewer describes the new Base/Overage bars as non-transparent and a costly query draining 3/4 of the Base bar and 1/8 of the monthly Overage bar, calling it “worse than any online service” they’ve used; there is no middle tier between $8 and $50, which multiple users in feedback threads ask for; a Trustpilot reviewer reports being charged after cancellation with no subscription showing and the contact button broken; it is hosted and not self-hostable, so data routes through T3 and external providers; the small Trustpilot base (4 reviews in 12 months) means reliability signals are thin.
Who is it for?:Users who want one fast UI to compare and switch between frontier models; developers who want to iterate on prompts and coding ideas before moving to a repo (not for agent orchestration—use T3 Code for that); power users willing to bring an OpenRouter key for cost transparency; people who prefer a hosted product over self-hosting an open-source alternative.