OpenRouter

What it is

OpenRouter is an online marketplace and gateway for large language models (LLMs). A developer does not need a separate account with every AI company whose models they want to use: they send a request to OpenRouter, the platform routes it to a model from one of the providers it works with, and the answer comes back. OpenRouter says it started in early 2023 as the first LLM marketplace, and that it has since grown into the largest and most popular AI gateway ^[1].

The two words describe different jobs. A marketplace is a place where many sellers offer their goods side by side and buyers compare them; a gateway is the single door through which traffic enters and leaves, taking care of routing and retries on the way. OpenRouter does both: it lists models from many providers, and it sits between the application and those providers.

Where it came from

The company states its mission as "increase AI neurodiversity" — building a transparent marketplace for AI that works everywhere without model lock-in ^[1]. Model lock-in is the situation an application ends up in when it is written around one vendor's model and interface: the code, the prompts and the evaluation all assume that model, so moving to a different or cheaper one later means substantial rework. A marketplace that routes to many providers is meant to remove that constraint, and the company puts its pitch in terms of better prices, uptime and enterprise-grade reliability ^[1].

How it works

Three features on the homepage describe the mechanics. The first is a single unified interface with OpenAI SDK compatibility: developers keep the calling conventions of that widely used software library and point them at this platform, instead of writing a separate integration for each provider ^[2]. The second is fallback: if the provider behind a chosen model is unavailable or declines the request, the platform can retry with another provider serving the same model, which raises availability ^[2]. The third is edge deployment, running the routing layer close to users so that latency stays low ^[2].

The scale the company reports is 700T+ tokens a month, more than 10 million global users, more than 80 providers and more than 600 models ^[1].

Funding

On 28 May 2026 OpenRouter announced a $113 million Series B led by CapitalG, with participation from NVentures, ServiceNow Ventures, MongoDB Ventures, Snowflake Ventures, Databricks Ventures, AMP PBC and Pace Capital, alongside existing investors Andreessen Horowitz and Menlo Ventures ^[2]. The press section of the company's about page lists coverage of a $40 million raise as well as of the $113 million Series B ^[1].

Where it sits

The catalogue keeps changing: similar tasks are served by several models with different prices and speeds, and a single provider having a bad day can take down a whole product path. An aggregation layer has a place in that picture, because it holds model choice, routing and failover in one spot rather than in every application. OpenRouter defines its value by the absence of lock-in — a marketplace that works everywhere ^[1]. Its published token volume and user count are the company's own measure of how far that argument has carried ^[1], and the two funding rounds on its press page mark the stages it has passed through so far ^[1].