We are raising a $750,000 pre-seed round

AI Stupid Level measures AI models on a fixed schedule, shows when one gets worse behind the same name, and routes each request to the model that measures best for it. This round takes it to market. AI model providers cannot invest.

Talk to us about investing Meet the team

Our figures and the terms are shared with investors privately, on request.

$750k: pre-seed round, with a minimum ticket of $25,000
80%: of the round goes to go-to-market, 20% to the benchmark and the router
$0: from AI model providers, as revenue or as investment

What we do

We measure.: Coding, reasoning and tool-use benchmarks run on a fixed schedule across models from OpenAI, Anthropic, Google, DeepSeek, Moonshot and Z.ai, with an hourly canary on top. The method is public; the tasks stay private, so no one can train on them.
We catch change.: Each model is compared with its own past. When one gets worse behind the same API name, the drift monitor shows it.
We route.: The Smart Router is one OpenAI-compatible endpoint that sends each request to the model measuring best for it, with 15 routing modes. It runs on the customer’s own provider keys, at zero markup.

Why it matters

Companies building on AI models choose them from launch-day benchmarks and pay for them by the token, while a provider can change a model behind the same name without saying so. Which model is best for a task, and at what price, keeps changing as new models ship.

An independent, continuous record of how models perform, and a router that acts on it, answers both questions. It has to come from someone with no stake in the result, which no model provider can be.

Traction

Account, growth and revenue figures are shared with investors privately.

Paying customers: Subscribers on monthly and annual plans, from independent developers to companies. We do not name customers without their permission.
Distribution: Built into RAGFlow, Fabric, DeepChat and opencodex; a verified n8n node on n8n Cloud; packages for LangChain, Langflow and Dify; an MCP server in the official MCP Registry.
Partners: Official partner of World Summit AI.
Open: A public Data API, listed in public-apis, and an open-source web application.
Record: Our first benchmark ran on 8 August 2025.

How we make money

Subscriptions: Pro, Developer and Teams, billed monthly or annually (plans).
Enterprise: Contracted scope, with single sign-on, SCIM provisioning, an audit trail and invoicing.
Assessments: The workload assessment: a customer’s own tasks run against three candidate models, with a written recommendation.
Custom benchmarking: Continuous measurement of a customer’s own workload, built and run by us.
Data licensing: The underlying performance data, licensed only to organisations that are not model vendors.

The Smart Router runs on customers’ own provider keys at zero markup: we sell the decision, never the inference.

Independence

We do not take investment from AI model providers.

We measure their models and route our customers’ traffic between them. A provider that owned part of us would have a stake in every score we publish and every request we route, so this round is closed to them. The same rule covers revenue: no model provider pays us, and we have no affiliate links.

The round

Stage: Pre-seed
Raising: $750,000
Minimum ticket: $25,000
Terms: The equity on offer, the valuation and the instrument are shared with investors privately, together with our figures.
Not open to: AI labs and model providers (why).

Use of funds	Share	What it pays for
Go-to-market	80%	Taking the product to the developers and companies who choose and pay for AI models.
Benchmark and router	20%	Expanding and developing the benchmark and the Smart Router.

This page is for information only. It is not an offer to sell, or a solicitation of an offer to buy, any security. Any investment is made only on the terms of signed investment documents.

Team

Ionut Adrian Visan: Founder & CEO
Alexandra Chirilă, PhD: AI Evaluation & Epistemology: the design of our reasoning evaluations
Marius Răzvan Palimariu: AI Infrastructure Lead, with experience from IBM, NVIDIA and Nscale

Full bios on the About page.

Talk to us about investing

Tell us who you are and the ticket you have in mind. We reply ourselves, share the figures and the terms, and set up a call.