Data & Research
Nova processes billions of AI API requests each day. We share anonymized, aggregate insights with researchers, journalists, and policymakers to advance honest public understanding of how AI is actually being used at scale.
Our data philosophy
We don't sell prompt data
Nova is an API platform, not a data business. We never sell, license, or monetize individual request contents or user-identifiable data.
Aggregate insights only
Everything we share is anonymized and aggregated to the point where individual users, organizations, or requests cannot be identified.
Public benefit focus
We prioritize data partnerships that advance policy, academic research, and journalism — not competitive intelligence or commercial targeting.
Institutional partners
Organizations that have used Nova data in published research or reporting.
NIST / CAISI
GovernmentTracked 97.5 million API requests over four weeks to study enterprise AI adoption patterns and model reliability.
MIT & Boston University
AcademicConstructed economic models of developer AI spend and switching costs using Nova usage data.
Andreessen Horowitz
ResearchCollaborated on the State of AI 2026 report analyzing open vs. closed model market dynamics.
Brookings Institution
PolicyCited Nova usage data in reports on AI workforce impact and geographic concentration of AI development.
Reuters
MediaUsed Nova's model traffic data to report on the rise of Chinese-developed AI models in global developer markets.
Cisco Systems
IndustryReferenced Nova's task distribution data in enterprise AI readiness and security research.
Available datasets
What we share — and what we don't.
Model market share
Weekly token share by provider and model family, broken down by modality and task category.
Task distribution
Aggregate distribution of API requests across task categories: coding, writing, research, roleplay, etc.
Geographic traffic
Token volume by region and country (account registration region, not IP geolocation).
Model switching patterns
Aggregate data on how developers migrate between models following price changes or capability releases.
Agentic usage trends
Tool-calling rates, average context lengths, and multi-turn session depth over time.
Raw prompt data
Individual request contents, user identifiers, or any data that could identify individuals or organizations.
Public resources
Request data access
Researchers and journalists can request anonymized dataset exports and custom cuts.
We typically respond within 2 business days. Academic and non-profit requests are prioritized.
research@nova.ai →