Weflayr
Category: AI Infrastructure
Paris-based AI gateway that routes each LLM request to the best model, based on benchmarks run against the customer's own production data. Weflayr was founded in 2026. The company is led by Romain Mazeries. Based in Paris, France. Team size: 1-10. Total funding raised: $2.5M. Latest round: Pre-Seed. Key investors include 13books Capital, Founders Future, Kima Ventures, Plug and Play, Station F Ventures, Roxanne Varza.
- Founded
- 2026
- Headquarters
- Paris, France
- Team size
- 1-10
- Total funding
- $2.5M
Value proposition
Gives AI companies control over model evaluation, selection and routing. It benchmarks models on the customer's real production calls, routes each request to the best approved model within the customer's rules, and ties inference costs to margins by customer and feature. The first deployment reported over 55% lower inference costs at equivalent quality.
Products and solutions
Weflayr AI gateway: benchmarking on customer data (replayed production calls across models), a routing engine that matches each request to similar benchmarked calls, custom routing rules that use business context such as customer plan and AI margin, one endpoint across providers, BYOK or managed keys, automatic fallbacks and rate-limit handling, cost and AI-margin analytics. Currently in private beta with a waitlist.
Unique value
There is no best model, there is a proven mix. Routing decisions are validated by benchmarks on the customer's own data and shaped by customer business context, instead of generic benchmarks or black-box routing.
Target customer
AI-native software companies and AI builders running LLM inference in production, in Europe and the US.
Industries served
AI infrastructure, SaaS and AI-native software companies
Technology advantage
Routing driven by production data: each request is matched to previously benchmarked similar calls. Routing is linked to business context (customer, plan, margin). The team brings payments-routing experience from Mangopay. EU-built, with SOC 2 Type I achieved, SOC 2 Type II and ISO 27001 in progress, and GDPR sub-processors in the EEA.
How they differentiate
Competing gateways mostly route without knowing how models perform on a given workload. Weflayr benchmarks models on the customer's own traffic, lets teams approve trusted models and regional deployments, and applies margin-aware rules. It positions itself as an independent, EU-based intelligence layer.
Main competitors
OpenRouter (acquired by Stripe), Not Diamond, Martian, also Portkey and LiteLLM
Key partnerships
Incubated in the Station F F/AI program, Plug and Play (investor)
Major milestones
2026: founded in Paris and joined the Station F F/AI program. 2026-10: raised a $2.5M pre-seed led by 13books Capital. SOC 2 Type I achieved. Private beta, with public launch planned soon.
Growth metrics
The first deployment showed over 55% lower inference costs at equivalent quality, and 15 percentage points higher relative quality at the same cost than a frontier model. The company is onboarding its first customers in private beta.
Market positioning
Early-stage European challenger in AI inference routing and AI gateways, positioned as an independent control plane for inference cost, quality and margins.
Geographic focus
Europe and United States
About Romain Mazeries
Ex-CEO of Mangopay, where he scaled the payments infrastructure business; about 15 years in payments, including payment routing systems.
Latest news about Weflayr
More AI Infrastructure companies
Official website: https://weflayr.com