Weflayr is building a next-generation AI gateway that gives AI companies full control over their routing strategy, optimizing AI quality, cost, and margins through benchmarking.
PARIS, FRANCE, 22 October 2026 – Weflayr, a French AI infrastructure startup founded by former Mangopay CEO Romain Mazeries and Noé Achache, former Head of AI Engineering at Theodo Data & AIe, today announced a $2.5 million pre-seed funding round led by 13books Capital, with participation from Founders Future, Kima Ventures, Plug and Play, Station F venture, Roxanne Varza (General Catalyst scout), and several angel investors. Weflayr is also part of Station F’s F/AI program.
The funding will be used to expand the engineering and AI research team, accelerate product development and launch go-to-market activities in Europe and the United States.
The AI model market is exploding but the infrastructure hasn’t caught up
Hundreds of AI models now compete on quality, speed, and cost, with new releases constantly reshaping the competitive landscape. Yet most AI companies still rely on a handful of models for the majority of their workloads.
The reason is simple: quality is non-negotiable. Different models excel at different tasks, but switching between them without compromising quality remains a major challenge.
Today’s AI gateways can route requests between models. But without knowing how those models actually perform on a company’s own workloads, they can’t reliably determine which one will deliver the best results.
For AI companies, this creates a fundamental dilemma: how do you optimize inference costs without putting the quality of your product at risk?
Weflayr believes you shouldn’t have to choose.
From black-box routing to an intelligent AI control plane
Weflayr puts companies in control of their routing strategy. Rather than relying on generic model rankings, Weflayr lets AI teams benchmark models against their own production workloads based on their quality and cost objectives, then approve the models they trust. Weflayr then dynamically matches incoming requests to similar previously benchmarked calls, routing each request to the model that performed best within those predefined boundaries.
Every decision also takes into account the business context of the request: the customer it serves and the margin attached to it.
First deployment has demonstrated promising results:
- Over 55% lower inference costs while maintaining equivalent quality.
- 15 percentage points higher relative quality at equivalent cost compared with using a frontier model.
Weflayr also connects AI consumption to business context, providing visibility into costs and margins across customers, features, and workflows. The ambition is to move beyond traditional AI gateways and become the intelligence and optimization layer between AI applications and model providers.
Because Weflayr sits in the path of production AI traffic, security is built in from day one. The company has completed its SOC 2 Type I audit and is working towards SOC 2 Type II and ISO 27001 certification.
Built by founders who have scaled critical infrastructure
Romain Mazeries, former CEO of Mangopay, built and scaled a payment infrastructure that processes billions of euros for marketplaces and platforms across Europe. Noé Achache led AI engineering at Theodo Data & AI, where he deployed production AI systems at scale for large enterprises.
“In the last 15 years, I have seen the payment landscape evolve toward very sophisticated routing systems. I am convinced that AI is following the same path, but at a much greater speed. AI builders want control and far more granularity in their routing decisions, and this is what Weflayr is about.” said Romain Mazeries
“Bridging the gap between evaluation and routing opens the door to a completely new way of building AI products, especially as we scale to the full harness. Weflayr is about defining this new category.” said Noé Achache
Mike McFadgen, Partner at 13books Capital commented: “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. Romain and Noé bring a rare combination of infrastructure scaling experience and deep AI expertise. Together, they have what it takes to build a category-defining company, and we’re proud to lead this round.”
Accelerating product development and international expansion
The $2.5 million investment will support Weflayr’s engineering and AI research efforts, accelerate product development, and expand its go-to-market activities across Europe and the United States. Weflayr is currently onboarding its first customers as part of a phased product rollout, with broader availability planned in the coming weeks.
Its long-term ambition is clear. Become the default control plane for AI inference, giving every AI company the power to continuously optimize model performance, cost, and margins.
