● Several models in one chat · projects · agent mode

Every AI model in one chat

Claude, GPT, Gemini, Grok, Kimi, DeepSeek and dozens of open models. They search the web, create files, generate images, talk, and understand photos and video.

No card required. Free plan forever.
ClaudeGPTGeminiGrokKimiDeepSeekGLMMiniMaxMistral QwenGemmaPhi-4GraniteMistral NeMoDeepSeek V2 Lite
Compare three open coding models under 15B and ask me which one we go with
Searched the web · 8 sources: Hugging Face, Reddit, GitHub
ModelSizeContextStrength
Qwen2.5 Coder 14B14B128Kbest code quality in its class
Phi-4 14B14B16Kreasoning and math
Qwen2.5 Coder 7B7B128Kspeed
A Qwen2.5 Coder 14B — quality over speed
B Qwen2.5 Coder 7B — fast autocomplete
C Phi-4 — more logic, less code

Everything you expect from a chatbot

Models don't just reply with text — they use tools when it helps.

Teams of agents

Give each agent a name, a model and a role — Planner, Coder, Reviewer — and watch them work on your task in a live team chat.

Deep Research

The model plans, reads dozens of sources and hands you a report with links to every one.

Run HTML in the site

Press Run on a page in your project and it opens in an isolated window right here.

Memory you control

The model suggests what to remember — and saves it only after you say yes.

Teams and credit packs

Share projects and agents with colleagues from $20 per seat, or top up with a one-time pack from $3.99.

Deep research

Web, Wikipedia, Reddit, Hugging Face, GitHub, arXiv, Stack Overflow — with links to every source.

Files in answers

Project code, documents, tables — as cards with a Download button, or all at once as a .zip.

Images and photos

Image generation plus real photos on the topic from openly licensed sources.

Sees photos and video

Every model understands images, video and audio — even models that can't see on their own.

Voice

Dictate instead of typing and listen to answers read aloud.

Polls and decisions

The model offers options as buttons — “A, B or C?” — and continues with your choice.

Reasoning depth

From instant answers to Max. See how the model thinks and what each answer costs.

Code sandbox

Models run the code they write, install libraries, read the errors and fix them before answering.

Several models at once

Add two to four models to one chat: they take turns and build on each other, or answer side by side so you can compare.

Agent mode

Hand over a whole task: the model makes a plan, writes and runs code, fixes errors and reports what it did.

Works in your coding tools

Connect Codex CLI, Claude Code, Cline or opencode with one API key and use any model of your plan on your own projects.

Projects

Keep chats, files and instructions together. Every chat in a project knows its files and shares one sandbox folder.

Privacy

Temporary chats, one-click export and deletion. Open models run on our own servers.

Pricing

Start free. Pay only when you need flagship models and higher limits.

Compare all plans and engines →