Reasoning that shows its work
Watch Nova plan, branch, and self-correct in real time — transparent chains of thought you can inspect, audit, and trust.
Frontier intelligence · Nova 3
The AI that reasons, writes, and ships alongside you.
One frontier model for text, code, and vision — with a million-token memory and answers you can actually trust.
Powering the teams building what comes next
Vercel•Linear•Ramp•Notion•Retool•Framer•Perplexity•Anthropic•
Vercel•Linear•Ramp•Notion•Retool•Framer•Perplexity•Anthropic•
Nova turns a blank page into finished work, a vague idea into a shipping product, and a hard question into an answer you can defend. It does not just autocomplete — it thinks, plans, and follows through.
Built to do the hard part.
Every capability, one API. Nova reasons across modalities and keeps its footing on your longest, messiest problems.
0
tokens of context — entire codebases and case files, held in mind at once.
Sub-second latency
Streaming responses in ~120 ms. Fast enough to feel like a thought.
From prompt to plan
Ask in plain language. Get structured, ready-to-ship output back.
Yours, privately
SOC 2 Type II. Zero training on your data. Deploy in your own cloud.
One model.
Every modality.
Drag through everything Nova can do —→
Text
Long-form writing, analysis, and synthesis that stays coherent across a million tokens.
Code
Writes, refactors, and debugs across whole repositories with full-project awareness.
Vision
Reads screenshots, diagrams, and documents — and reasons about what it sees.
Voice
Natural, low-latency speech in and out, in forty-plus languages.
Agents
Plans multi-step work, calls your tools, and finishes the job unattended.
0
Platform uptime
0
Median latency
0
Languages
0
Requests / day
Loved by the people who ship.
“Nova replaced three tools and a weekly meeting. It reads the whole codebase and just knows.”
Daniel OkonkwoStaff Engineer, Ramp
“The reasoning is the difference. It doesn’t guess — it shows me why, and it’s right.”
Priya NairHead of AI, Linear
“We shipped a feature over a weekend that would have taken a quarter. Genuinely absurd.”
Marcus FeldFounder, Cadence
Start free. Scale when it’s working.
No credit card to begin. Usage-based when you grow.
Explore
$0/mo
- 50K tokens / day
- Text & code
- Community support
- 1 seat
Most popular
Pro
$29/mo
- Unlimited tokens
- All modalities + Vision
- 1M context window
- Priority latency
- 5 seats
Enterprise
Custom
- Private deployment
- SOC 2 + SSO
- Dedicated capacity
- Solutions engineer
Questions, answered.
Is my data used for training?
Never. Nova does not train on your prompts, files, or outputs. Enterprise plans add private, single-tenant deployment inside your own cloud.
How big is the context window?
Up to one million tokens on Pro and Enterprise — roughly an entire codebase or a stack of case files held in memory at once.
Which languages are supported?
Forty-plus natural languages and every major programming language, with the same reasoning quality across all of them.
Can I self-host?
Yes. Enterprise customers can deploy Nova in a dedicated VPC with SSO, audit logs, and SOC 2 Type II controls.
Build something impossible.
Ten minutes from now you could be shipping with the best model you’ve ever used.
Start building free