How Depaza works
One choice. A fleet of EU models behind it.
You pick a tier. Depaza matches every request to the model built for the job — fast answers, deep reasoning, finished documents, vision, voice — all hosted in the EU, none of it crossing the Atlantic.
Pick a tier, not a model number.
Most tools chain you to one model and make you babysit its version number. Depaza gives you three simple tiers. Choose how much horsepower you want — we handle which engine runs underneath.
Depaza Core
RecommendedThe everyday default
The best balance of intelligence, speed and tool use. The right answer for most work, with web search and document tools built in.
- Best all-round quality and speed
- Web search and tools included
- Available to everyone, including Free
Depaza Max
For depth
Our flagship tier for deep analysis, complex reasoning and the most demanding work — when getting it exactly right matters more than getting it instantly.
- Deep analysis and hard problems
- Long, structured output
- On paid plans
BOSS
Maximum candor — advanced
The most direct, least filtered tier, for advanced users who would rather have a blunt answer than a careful one. Use with judgement.
- Raw, direct tone
- Fewer guardrails
- Advanced users, paid plans
Free runs Core at a daily limit; Max and BOSS unlock on paid plans. Switch tier per conversation — your history and files come with you.
The right model for the job
Behind one tier, a team of specialists.
A tier is your dial for horsepower — not one model doing everything. Under the hood, Depaza routes each step to a model built for that exact task, and stitches the results into one answer. You never pick which engine reads your PDF or drafts your deck. We do.
Conversation
Fast, grounded answers that stream as they are written — tuned for everyday questions and tool use.
Deep reasoning
In Expert mode a second model reviews and challenges the first before you ever see the answer — more like a panel than a single take.
Documents
A drafting model purpose-matched to native PowerPoint, Word and Excel — real files, not text to paste.
Vision
A dedicated EU vision model reads images, screenshots and scanned PDFs so the assistant can "see" what you upload.
Voice
EU speech-to-text turns recordings and dictation into clean text inside the chat.
Live web research
When a question needs fresh facts, Depaza searches the live web, reads the pages and cites them — no invented dates.
Always improving. Never locked in.
Because you choose a capability tier instead of a brand-name model, we can keep upgrading the engines underneath you — your prompts, projects and workflows keep working.
Quietly upgraded
We continuously evaluate new models and swap in better ones behind each tier. You get the improvement without chasing version numbers or rewriting prompts.
Open models, no lock-in
Depaza runs on open models, not one vendor's black box. Use our hosted EU models, or self-host your own with the Depaza Code CLI (Ollama, vLLM or local) — same tools, same workflow.
Documented for compliance
We document our provider role and the models behind every tier, and share that documentation with your compliance team on request — ahead of the EU AI Act obligations.
EU First
Every model runs in the EU.
EU compute, EU storage, EU model training — your prompts and files never cross the Atlantic. No US provider ever processes your conversations or documents.
EU compute and storage
Every model — chat, documents, vision and voice — runs in EU regions, alongside your database, storage and backups. No US provider ever processes your prompts or files.
Training stays in the EU
We may use your prompts, conversations and files to improve Depaza's own EU-hosted models — and that training, processing and storage stays in EU regions, never across the Atlantic. Enterprise plans are excluded from training.
Try it on your own work.
Create an account in 30 seconds. Ask a hard question, upload a file, generate a document — and let Depaza pick the right model for each one.