Neurometric raises $4M and launches automated token engineering platform for agentic AI
What's the deal? New York-based NeurometricDealroom has a profile for this one. Try Dealroom → has launched an automated "token engineering" platform and revealed $4 million in funding closed in early 2026.
The startup routes each AI task to the cheapest model that can do the job, tweaks prompts as needed, and builds a custom small language model when no existing option fits.
The round drew BetaworksDealroom has a profile for this one. Try Dealroom →, ex-Ante, Everywhere.vcDealroom has a profile for this one. Try Dealroom →, EncodedDealroom has a profile for this one. Try Dealroom →, VermillionDealroom has a profile for this one. Try Dealroom →, AbstractionDealroom has a profile for this one. Try Dealroom →, and Mu VenturesDealroom has a profile for this one. Try Dealroom →, plus angels Jason CalacanisDealroom has a profile for this one. Try Dealroom → and HubSpot chief technology officer Dharmesh ShahDealroom has a profile for this one. Try Dealroom →.
Why now? Companies are moving AI agents from pilots into production, where a single workflow can trigger dozens of model calls.
Many still send every task to a costly frontier model, even when a smaller one would match or beat it. "Every model call is also a pricing decision, and those decisions compound across an agent's workflow," said chief executive officer Rob May.
Neurometric brings model routing, small language model creation, and a marketplace of pre-trained models into one platform. In early tests, its routed models beat frontier models by up to 20 points on accuracy while cutting cost and latency.
What could go wrong? The pitch rests on a market that shifts constantly, with new models, prices, and performance benchmarks landing weekly.
"Things change so fast a human token engineer can't keep up," May said. That speed is Neurometric's selling point, but it also means rivals — and the frontier labs themselves — can move just as quickly.
The signal: Backing from Betaworks, an early-stage specialist, points to investor appetite for the infrastructure layer that makes agentic AI economically viable rather than the agents themselves. Neurometric, still an early-stage company, is positioning its automated inference orchestration as the cost-control tool businesses will need as multi-model AI systems scale.
Read more: StreetInsider
Image credit: Robert Scoble
