Oyren Chat: several models, one question
One question, several answers.
One question goes to up to five models at once, and each answer is signed by its model. Free with an own Google, OpenAI, Anthropic or DeepSeek key, per token otherwise. Chats, notes, PDFs, boards and the agents' computers share one workspace.
Seats
Several models answer the same question.
A conversation seats up to five models, and each answer is signed with the model that wrote it. One model's blind spot becomes a bug. A second opinion is one more seat.
- Every answer is attributedThe model's name and provider sit above what it wrote.
- All at once, or in turnIn order, each seat can see what the earlier ones just said.
- Billed on real token useTwo answers is about twice the tokens, and the page says so before sending.
- Models from eight vendorsAnthropic, OpenAI, Google, DeepSeek, Qwen, Moonshot AI, Z.ai, MiniMax, as the catalog lists them.
- Custom agents take a seatAn agent with its own instructions and model joins as an extra seat.
A dedicated computer for agents.
Oyren gives agents a work environment that runs 24/7. Inside it, agents write notes, read PDFs, pick up cards from boards, write code and use a browser.
Always on
Agents keep working in the cloud computer after the laptop closes.
Bypass permissions, kept inside
Agents run with bypass permissions, so unattended work never waits on a prompt. An accidental deletion stays inside that computer, away from personal files.
Cloud now, spare devices next
Cloud computers run agents today. Raspberry Pi and old Linux laptops come next, and Macs and Windows PCs are planned.
Documents in the workspace
It answers out of the files.
A file can be pinned to the thread or named with @. The reply carries the passage it used, so it can be checked against the document.
- Pin or @-mentionThe PDFs, notes and scans the workspace already holds.
- Answers carry the quoteThe line it read, and the file it came out of.
- Web search is opt-inA checkbox in the composer, 15 credits a search when it is ticked.
“Limit Friday to 24 boxes” from idea.md
Image generation
Images generate in the same composer.
The composer switches to Generate with one image model from the catalogue. The picture comes back in the thread, already saved as a workspace file.
- One model, picked per imageGeneration runs on that model alone, even in a thread full of seats.
- Images come back straight awayNo job to watch. The result lands in the reply.
- Credits or an OpenRouter keyOyren credits, or an OpenRouter key and Oyren charges nothing.
Video generation
A video is a job that runs in the background.
Video takes minutes, so it runs as a job instead of a reply. The card in the thread shows where the render has got to, and the thread picks the job up again after the tab closes.
- Progress in the threadSubmitted, rendering, then saved into Files.
- Priced by the secondHeld before it starts, settled on the seconds actually rendered.
- The file stays in the workspaceIt lands in the workspace next to the documents the thread was reading, with a download.
One thread, several models, one workspace.
The workspace is free to keep. Payment applies when a model runs, and the price is on screen before it does.