Weflayr, a French startup that builds software for choosing which AI model handles each request, has raised $2.5m in pre-seed funding led by 13books Capital. The company announced the round on 22 October 2026.
Founders Future, Kima Ventures, Plug and Play and Station F Ventures also invested, along with Roxanne Varza, director of Station F, and a group of angel investors. Maddyness reported the round in euros as €2.2m.
The company did not disclose a valuation.
“AI inference is rapidly becoming a critical infrastructure layer for the next generation of software companies. As the market evolves and consolidates, we believe there is a massive opportunity for an independent intelligence layer that gives AI companies full control over quality, performance, and economics.”Mike McFadgen, Partner, 13books Capital
What Weflayr does
Companies that build AI products can choose among hundreds of models that differ in quality, speed and price, but most rely on a handful because they cannot easily tell which model performs best on their own tasks. Weflayr argues that existing AI gateways, which pass requests between models, make those choices without knowing how each model performs on a customer's real workload.
Its product lets customers benchmark models against their own production calls, set quality and cost goals, and approve the models they trust. Weflayr then matches each incoming request to similar calls it has already tested and sends it to the best approved model within those limits. It also links each inference decision to costs and margins by customer, feature and workflow.
The company said its first deployment cut inference costs by more than 55% at equivalent quality and delivered 15 percentage points higher relative quality at the same cost as a frontier model. These results are the company's own and were not independently verified. Weflayr said it is onboarding its first customers and plans wider availability in the coming weeks.
The founders
Weflayr was founded by Romain Mazeries, the former chief executive of payments company Mangopay, and Noé Achache, previously head of AI engineering at Theodo Data & AI. The company compares its approach with the way smart routing changed payment processing. It is incubated at Station F in Paris, according to Maddyness, which described the product as a routing layer between a company's applications and AI models.
What the money is for
Weflayr said the money will be used to grow its engineering and AI research teams, speed up product development and bring the product to more AI companies in Europe and the US.
Mike McFadgen, a partner at 13books Capital, said in the announcement that AI inference is becoming critical infrastructure for software companies and that there is room for an independent layer giving those companies control over quality, performance and cost. This is the company's first funding round, according to Maddyness.
A busy week for French tech
Weflayr was one of seven French startups whose rounds Maddyness counted in its weekly round-up, which put the total raised that week at close to €271m. Most of that came from a single deal, a Series E for the AI planning software company Pigment. Two other rounds in the list were also aimed at AI infrastructure and agents: Reduck, which raised €2.7m for web automation tools used by AI agents, and Webyn, which raised €7m for AI agents for e-commerce.
Company website: weflayr.com