Neurometric AI announced that it has launched its automated token engineering platform and raised $4 million in funding. The round included participation from Betaworks, ex-Ante, Everywhere.vc, Encoded, Vermillion, Abstraction, and Mu Ventures. Angel investors included Jason Calacanis, co-host of the All-In Podcast, and Dharmesh Shah, CTO of HubSpot.
The company is focused on helping businesses control the cost and performance of agentic AI workloads. The funding closed earlier this year.
Neurometric said companies moving AI agents from experimentation into production often face higher costs because a single workflow can generate dozens of model calls. Many businesses continue sending each task to a frontier model, even when a smaller and less expensive model could deliver the same or better result.
Neurometric evaluates those calls individually, modifies prompts when needed, and routes each task to the most cost-effective model that can meet the required performance threshold.
When no existing model fits a customer’s requirements, Neurometric can create a purpose-built small language model for the task. For simple high-volume workloads, the platform automatically generates specialized small language models to optimize speed and cost.
The company’s platform brings model routing, small language model creation, and access to a marketplace of pre-trained task-specific small language models into one system.
Neurometric’s Task Endpoint Manager evaluates incoming requests against continuously updated model performance and pricing data. It then routes each task based on the customer’s accuracy, cost, and latency requirements.
The platform’s Auto-SLM Creator builds and serves small language models when existing models do not meet those requirements. The company’s SLM marketplace also allows customers to access models developed for common and recurring workloads.
Neurometric said early customer engagements have shown that models routed or created through its platform have beaten frontier models by as much as 20 accuracy points while reducing cost and latency compared with using frontier models for the same work.
Neurometric plans to use the funding to expand its engineering and AI research teams and add more optimization tools to its core platform.
The company is positioning token engineering as a new discipline for determining how each task within an AI workload should be completed based on quality, cost, and speed. Unlike prompt engineering, which focuses on improving instructions given to a model, token engineering determines which model should receive a task and whether a specialized model should be created to handle it.
Neurometric’s automated token engineering platform is available now.
KEY QUOTES:
“Companies have spent the past year proving that AI agents can perform increasingly complex work. Now they have to prove the economics still make sense when those agents are operating at scale. Every model call is also a pricing decision, and those decisions compound across an agent’s workflow. Token engineering gives companies a way to control that cost without sacrificing quality.”
“The number of available models is growing too quickly for companies to evaluate every option by hand. And the number of tools and techniques to improve them are growing even faster. Things change so fast a human token engineer can’t keep up. That decision needs to be automated and continuously reevaluated as the market changes.”
“Frontier intelligence will keep getting less expensive, but companies will also consume far more of it. The winners will not be the businesses that simply buy the most tokens. They will be the ones that know where advanced intelligence creates value and where a smaller model can do the job just as well.”
Rob May, CEO of Neurometric
“Companies need to know where frontier-level performance is worth paying for and where a smaller model can deliver the same result at a fraction of the cost. That discipline will determine whether agentic AI can move from promising pilots to a business model that scales.”
Calvin Cooper, COO of Neurometric
“Neurometric is tackling one of the most pressing problems in the AI ecosystem today. The team has a unique mix of AI talent and systems engineering experience that positions them well to take on this task and help companies optimize their token spend at multiple layers of their infrastructure.”
Alex Benik, Investor at Encoded
