OpenRouter

OpenRouter is an AI model gateway — a platform that exposes hundreds of large language models from many different providers through a single, OpenAI-compatible API, with one account, one credit balance and automatic routing between providers. Developers use it to compare and mix models such as GPT, Claude, Gemini, DeepSeek and open-weight alternatives without maintaining separate integrations for each vendor. In August 2026, the payments company Stripe announced an agreement to acquire OpenRouter in a deal reported at approximately US$7.5 billion.^[1][2][3]

History

OpenRouter was founded in 2023 in New York by Alex Atallah, a co-founder of the NFT marketplace OpenSea, and launched as model proliferation made it increasingly impractical for developers to integrate every provider individually. The platform grew alongside the multi-model era, becoming a common way to benchmark and switch between models, and expanding its catalogue past 500 models from more than 80 providers. It raised a US$12.5 million seed round led by Andreessen Horowitz in February 2025 and a US$28 million Series A led by Menlo Ventures in April 2025, followed by a US$113 million Series B led by CapitalG in May 2026 at a reported US$1.3 billion valuation. Attention on the platform intensified in July 2026 when reporting based on its usage data found that Chinese-origin open models had captured 46 per cent of United States enterprise token usage on the service. On 19 August 2026, Stripe announced it had agreed to acquire OpenRouter, with reports valuing the deal at more than US$7 billion.^{[3][4][5][6][7]}

Key Concepts and Technology

OpenRouter normalises provider APIs into a single interface: a developer sends an OpenAI-style request naming a model by slug, and the platform routes it, translates parameters, and returns a standardised response, so switching models means changing one string. Routing features include automatic failover when a provider is unavailable or throttled, an auto-router that selects a model based on the task and a cost-quality trade-off, preferred-vendor selection, and pass-through pricing — provider token rates plus a platform fee on credit purchases, rather than a markup on tokens. The platform also supports bring-your-own-key configurations, privacy controls such as zero-data-retention routing and regional routing options for businesses, and budget and spend controls; its free tier offers access to a set of free models with daily request limits. Usage analytics and activity logs let teams see which models requests actually reached. The main criticisms concern the dependency of a central intermediary between applications and model providers, and the caveat that privacy characteristics vary per provider and must be checked for each route.^[1][2][8]

Applications and Impact

Developers use OpenRouter to prototype quickly, to run evaluations across many models with one integration, to reduce the engineering cost of switching vendors, and to give products resilience by failing over between providers. For startups, it lowers the barrier to accessing frontier models from multiple labs without multiple accounts and billing relationships; for larger organisations, routing and spend controls provide a degree of governance over model usage. The platform's usage data has also become an informal industry indicator — its statistics on the growing share of open-weight and Chinese-origin models were widely cited in 2026 as evidence of a shifting model market. The Stripe acquisition, if completed, would fold model routing into payments infrastructure, extending the gateway model into billing, metering and monetisation for AI products.^[6][7][8]

>See Also

• LLM Routing
• Large Language Models
• Open-Weight Models
• Hugging Face
• Together AI