You can choose how hard it thinks!
Five effort levels, from a quick Light answer to Ultra, which drafts, reviews its own work like a strict editor, then answers. You see the cost before you send.
διανοητικός · thinking step by step
AI that shows its work.
Where thought takes form.
Turn up how hard it thinks. Watch the code it writes run right beside the chat. Pull your whole team into one conversation. Free to start.
See it in action
48 seconds
What's different?
Watch it build something...
Dianoetic Code writes complete projects, not snippets, and runs websites and games live next to the chat in a locked sandbox. Ask for a game, play it ten seconds later.
Think and Work together!
Invite your team into one chat. Everyone sees prompts and answers as they're written, and each person's prompts use their own plan.
How it works
No single model is best at everything. Dianoetic is designed as an orchestration layer: it decides how a problem should be approached, then brings in the right model, tool or source for each part.
-
Route
Model routing
Fast models for simple questions, stronger reasoning models when a problem calls for depth. Open models from several providers, with fallbacks when one is unavailable.
-
Reason
Effort levels
Five levels, from Light for quick answers to Ultra, which plans, drafts and reviews its own work before replying. You choose how much thought a question deserves.
-
Act
Tools and agents
Search, calculation, code and retrieval, used when they help and left alone when they don't.
-
Recall
Memory and knowledge
Context that carries across conversations, and retrieval over your own documents, so the system builds on what it already knows.
Products
- Dianoetic AI
- Conversation, research, writing and everyday coding help.
- Dianoetic Code
- Builds complete files and scripts, not snippets. Especially strong in C# and Unity.
- Dianoetic Images
- Creates images from a sentence, explains any picture you drop in, and converts, resizes and compresses image files.
The Dianoetic family
- Oriel 2.0
- Our strongest model, for deep reasoning and serious code.
- Fable 2.0
- Creative and general-purpose, for writing, ideas and conversation.
- Canto 2.0
- Fast and light, for everyday questions.
Where things stand
Dianoetic is being built in public, one layer at a time. Here's where things actually stand.
- Done
Foundation and orchestration. The site, accounts, and a server-side router across open model providers with automatic fallback.
- Done
Tools and effort levels. Plugins, uploads, images, and five effort levels you choose per message.
- Now
The 2.0 generation. Sharper routing, specialised instructions for Oriel, Fable and Canto, date awareness, and Ultra's draft-then-review pass.
- Later
Evaluation and our own data. Re-running benchmarks on 2.0, learning from ratings and beta tests, and fine-tuning when resources allow.
No papers yet. When we publish something, it will be because there's something worth reading.
For developers
The same system, through an API. One endpoint, five effort levels, and routing handled for you. Included with Thinker and Thinker Max.
curl https://dianoetic.tech/v1/chat \
-H "Authorization: Bearer $DIANOETIC_KEY" \
-d '{
"product": "code",
"model": "oriel-2.0",
"mode": "high",
"messages": [{ "role": "user", "content": "A C# rate limiter for ASP.NET Core" }]
}'