A studio for thinking with AI
Set your instructions once. Pick a model per question. Branch when an answer deserves another look. Everything stays in one place — including the answer you almost kept.
Live thread
Choose the right model for the next turn.
· Both labs, six minds ·
Practice
The useful part of AI work is rarely the first answer. It's the second attempt, the model switch, the context you keep carrying, and the branch you can return to when a better line of thinking appears.
oran.chat is the room around that loop. Instructions stay attached. Files travel with the thread. Answers split into branches. The model picker is always one turn away. Less ceremony — more attention on the work.
Instruments
Each feature carries one practical loop: instruct, ask, switch, branch, attach, continue. The studio handles what travels between turns.
Set your voice, your stack, your constraints, your preferences — once. They prepend every chat and follow every model switch. Your working style doesn't reset when you change advisors.
GPT-5.4 mini, GPT-4.1, o4-mini, Claude Opus 4.7, Sonnet 4.6, and Haiku 4.5. Pick per message and switch mid-thread.
A new answer becomes a branch off the same parent. The original stays available, and each path keeps moving until you decide which one was the work.
Drop in PDFs and screenshots. Vision-capable models see images directly, and document context stays in the conversation.
Pro adds cited web search, so fresh facts get checked instead of guessed.
10 messages a day on every model, including Opus 4.7. No card required. Upgrade when limits become the constraint, not onboarding.
In motion
Switch the model mid-thought. The prompt, files, and custom instructions stay attached while the answer becomes a branch.
You /
The point is not more chat windows. It is a better conversation graph.
At the desk
oran.chat is for people who already know what they want from AI and need a quieter place to do the work.
Keep a code-first system prompt on every chat. Switch to a larger model when the bug needs deeper reasoning. Attach screenshots, logs, and specs without restarting the thread.
Pin the voice once, then compare models against the same brief. Branch rewrites without losing the original angle or the notes that led there.
Drop PDFs in, ask with context, then use cited web search on Pro when the answer needs current facts. Keep interpretations in separate branches.
Pricing
Two tiers. The studio is in both. Pro lifts the limits and adds web-search citations.
Enough to settle in.
Quota resets at midnight local.
For the days you would rather not count.
Cancel any time, keep your history.
FAQ
OpenAI's GPT-5.4 mini, GPT-4.1, and o4-mini, plus Anthropic's Claude Opus 4.7, Sonnet 4.6, and Haiku 4.5. New models show up quickly after release.
Yes. Set a system prompt in settings and it prepends every chat across every model. Switch models mid-thread and the instructions follow.
Because the switching cost is real. oran.chat keeps GPT and Claude in one thread, one instruction set, branching, attachments, and Pro web search with citations.
10 messages every day, on every model including Opus. No credit card and no trial cliff. Quota resets at midnight local.
Unlimited messages on every model, web search with citations, larger context windows, higher attachment limits, priority routing, and early access to new models.
No. Conversations stay private to your account. OpenAI and Anthropic do not train on API traffic by default, and we do not use your content for model training.
Regenerate a reply and the new answer becomes a branch off the same parent. The original stays available, and each branch can continue from there.
Still curious? hello@oran.chat
The door is open
No credit card. 10 free messages every day on every model, including Opus 4.7. Custom instructions, branching, and attachments are all on the free tier.