💬
Smart auto-routing
We pick the best model for the task — DeepSeek for code, Gemma for vision, Llama for chat, FLUX for images. You just type.
Live · Free forever · No credit card
One chat, every AI model.
Omnivra AI auto-routes your prompt to the best free model — DeepSeek for code, Gemma for vision, FLUX for images, and live web search for fresh facts. You just type.
Powered by OpenRouter · Groq · Hugging Face · Tavily · Supabase
groq · llama-3.3-70b-versatile
You: what's the latest claude model?
Anthropic's newest is Claude Opus 4.7, released alongside Sonnet 4.6 and Haiku 4.5. Opus is the flagship for complex reasoning. [1][2]
2 web sources
Built for every kind of question
Most chatbots send everything to the same model. Omnivra picks a specialist and falls back when free tiers break.
🌐
Live web search
Tavily-powered search runs before the answer when you ask about the latest news, prices, or current events. Citations included.
🖼️
Vision Q&A
Upload any image and ask questions about it. Powered by Gemma 4 multimodal models.
🎨
Image generation
FLUX → SD3 → SDXL fallback chain so you always get an image even when free credits run low.
⚡
Provider fallback
Free LLM endpoints break constantly. Omnivra silently retries across Groq, OpenRouter, and Hugging Face so you don't see the failures.
🔒
Sign in & sync
Google or GitHub login. Conversations are stored in your own Supabase project with row-level security.
How does auto-routing work?
Every message is scanned for intent. Code snippets, math symbols, time-sensitive phrasing, image keywords, attached images — each triggers a different specialist model. When one provider is rate limited or out of credits, the router silently retries the next.
| Your prompt looks like | Routed to |
|---|---|
| "how do I fix this stack trace?" | DeepSeek V4 Flash |
| "prove the binomial theorem" | GPT-OSS 120B |
| "generate an image of a fox" | FLUX.1-schnell |
| photo + "what's in this?" | Gemma 4 31B |
| "latest AI news today" | Tavily → Llama 3.3 |
| "hi, how are you?" | Llama 3.3 70B |
Frequently asked questions
Is Omnivra AI free?
Yes. It routes through free tiers of OpenRouter, Groq, Hugging Face, and Tavily. You bring your own API keys (all have generous free tiers, no credit card required).
Which AI models does it use?
Llama 3.3 70B and OpenAI GPT-OSS 120B for chat and reasoning, DeepSeek V4 Flash for code, Google Gemma 4 for vision, FLUX.1-schnell for images, and any other model you wire into the router.
How is this different from ChatGPT?
ChatGPT uses one model for everything. Omnivra auto-picks the right specialist model per task and falls back across providers when free quotas run out — so you get better answers without managing accounts at five different AI companies.
Where is my data stored?
Your conversations live in your own Supabase project (free tier includes 500MB Postgres). Row-level security guarantees nobody else can read them. We never store them on our servers.
Can I self-host it?
Yes — the code is fully open. Backend is FastAPI, frontend is Next.js. Deploy frontend to Vercel and backend to Render in under 10 minutes.
Stop paying for five AI tools.
Get one chat that knows when to think, when to search, and when to draw.
Start chatting — free