Groq · Mistral · Gemini — routed automatically

One assistant. Every frontier model. Completely free. 

Nova reads each message and routes it to the model that handles it best — Codestral for code, GPT-OSS for reasoning, Gemini for vision, Llama for speed. You just type. No keys, no juggling tabs.

No credit card · No usage caps from us · History stays on your device

nova — new chatAUTO → Codestral
Write a debounce hook in TypeScript.

Seven models. One window.

Llama 3.3 70BGPT-OSS 120BMistral LargeCodestralGemini 2.5 FlashLlama 4 ScoutLlama 3.1 8BLlama 3.3 70BGPT-OSS 120BMistral LargeCodestralGemini 2.5 FlashLlama 4 ScoutLlama 3.1 8B
0

frontier models

0

providers, one chat

$0

cost to you

<0s

to first token

Capabilities

Everything you'd switch tabs for — in one place

Coding, reasoning, vision, and fast chat, each handled by the model that does it best.

Automatic model routing

Nova classifies intent on every message and picks the strongest free model for the job — no dropdown gymnastics. Override any time.

Built for code

Codestral-backed coding mode, syntax-highlighted blocks, and one-click copy.

Real reasoning

Hard problems route to GPT-OSS 120B with a transparent thinking trace.

Vision in

Drop in screenshots or diagrams — Gemini reads them natively.

Groq-fast streaming

Sub-second first tokens on Groq's LPU. Replies stream as they think.

Private by default

API keys stay server-side. Your chat history lives in your browser, not our database.

Stop picking models.
Start getting answers.

Join Nova and let the right model meet every message — free, forever.

  • No credit card
  • 7 models
  • Vision + coding
  • Private history