Vibeleaderboard
OpenRouter

router

Also indexed as open-router

OpenRouter

OpenRouter matters because it makes models and serving providers more interchangeable at the application boundary. That can improve resilience, cost control, and developer choice, and its traffic data offers a rare view of actual model use. The same intermediary position creates a trust problem: developers need to know which provider handled a request, what policies applied, and whether a routing decision changed model behavior or privacy exposure.2,4,5,11

Official site ↗Models in the Atlas →

Profile

Overview

A marketplace rather than a model lab

OpenRouter is an AI inference gateway and model marketplace founded in 2023 by Alex Atallah and Louis Vichy. It gives developers an OpenAI-compatible interface for models from many developers and serving providers, with centralized billing and a public catalog of price, context, modality, and provider information. OpenRouter is not primarily a foundation-model laboratory, even though routed or curated model identifiers can make it appear in model catalogs.1,7,6

Routing across models and providers

The service separates the model a developer requests from the provider that serves it. Its routing layer can load-balance among providers, order endpoints by price or throughput, fall back after failures, require parameter support, and restrict requests by data-retention policy. That creates a useful abstraction over a market in which the same model may be available from its developer, a cloud platform, and specialist inference companies with different prices and operational behavior.2,3,4

A view of real-world model traffic

OpenRouter's other important asset is market observation. It publishes model and provider activity derived from non-textual generation records such as model identifiers, token counts, timing, geography, latency, and tool-use metadata. A research report based on more than 100 trillion tokens describes the data as observational rather than representative of all AI use. The distinction is important because availability, price, and the platform's own user mix shape every usage ranking it publishes.5

An agreement to join Stripe

The company raised a $113 million Series B led by CapitalG in May 2026, after a $40 million Series A in 2025. In August, OpenRouter and Stripe announced an acquisition agreement. Independent reporting placed the value above $8 billion and interpreted the deal as giving Stripe a position in model selection and token economics. The agreement validates routing as strategic infrastructure, while making OpenRouter's provider neutrality, disclosure, and data-policy controls more consequential if the transaction closes.6,8,9,10

Notable contributions

  1. 01A unified multi-provider model marketplaceOpenRouter combined one compatible API, centralized billing, a model catalog, and provider competition in a developer-facing marketplace. It calls itself the first LLM marketplace; the broader contribution is independently visible in its funding, usage, and acquisition record.1,6
  2. 02Policy-aware provider routingThe gateway exposes provider order, fallbacks, price and throughput preferences, parameter support, regional handling, and zero-data-retention restrictions as routing controls rather than requiring separate integrations for each endpoint.2,4
  3. 03A large observational model-usage recordOpenRouter aggregated non-textual generation metadata into a dataset covering more than 100 trillion tokens, enabling analysis of model use across tasks, locations, prices, and time. It is a platform sample, not a census of the AI market.5