Stripe has agreed to amass OpenRouter, an AI model-routing platform that offers builders entry to a whole lot of fashions by means of a single interface. The deal provides mannequin choice and routing to Stripe’s present work round AI utilization and token-based billing.
OpenRouter helps greater than 400 fashions from over 80 suppliers, in accordance with Stripe. Slightly than requiring separate integrations with every mannequin supplier, builders can use OpenRouter to ship requests by means of one API.
Routing past mannequin alternative
The platform evaluates requests utilizing components together with process complexity, value, velocity, and reliability. It may well then direct every request to a mannequin suited to these necessities.
OpenRouter additionally handles a second layer of routing between suppliers serving the identical mannequin. Its documentation says prospects can prioritise endpoints based mostly on value, throughput, or latency, whereas setting necessities equivalent to most costs or minimal efficiency ranges.
The platform measures latency and throughput for particular person model-provider mixtures utilizing rolling efficiency information. This enables a request to be routed to an endpoint that meets specified price or efficiency standards quite than counting on a hard and fast supplier.
That separates two routing choices: which mannequin handles a request and which supplier endpoint serves it.
Supplier alternative also can have an effect on inference prices even when the underlying mannequin stays the identical. In a June 2026 instance, OpenRouter listed Llama 3.3 70B enter pricing at $0.10 per million tokens by means of DeepInfra and $1.04 by means of Collectively, whereas output pricing ranged from $0.32 to $1.04 per million tokens throughout the suppliers proven.
Routing can present failover when an endpoint turns into unavailable. OpenRouter says its system can transfer requests to different suppliers or fashions when it encounters issues together with supplier outages, fee limits, context-length errors, or moderation refusals.
Information-handling necessities also can type a part of supplier choice. OpenRouter lets customers prohibit requests to Zero Information Retention endpoints and forestall routing to suppliers that gather information or prepare on prompts, whereas enterprise prospects can request in-region processing within the US or EU.
Routing standards can due to this fact embrace mannequin functionality, supplier availability, processing location, latency, throughput, and price.
Multi-model infrastructure expands
Multi-model environments are already widespread amongst surveyed organisations. F5’s 2026 State of Software Technique report, based mostly on responses from greater than 1,100 IT decision-makers, discovered that 52% of organisations had been chaining or orchestrating a number of AI fashions, with respondents utilizing a median of seven fashions.
Menlo Ventures, an OpenRouter investor, reported a special measure of supplier behaviour in its 2025 mid-year survey. It discovered that 66% of builders upgraded fashions whereas staying with their present supplier, whereas 11% switched distributors.
OpenRouter can be one in every of a number of infrastructure suppliers including mannequin routing. Snowflake introduced dynamic mannequin routing for Cortex AI Gateway on August 18, with the function anticipated to enter personal preview.
Snowflake mentioned the system will assign requests in accordance with components together with high quality, velocity, buyer preferences, and price. Cloudflare presents Dynamic Routing in beta by means of AI Gateway, with guidelines masking mannequin choice, quotas, and fallbacks.
AWS offers Clever Immediate Routing by means of Bedrock, whereas Microsoft Foundry presents routing profiles that steadiness mannequin high quality and value. AWS and Snowflake each describe programs that may direct much less demanding workloads to smaller or lower-cost fashions whereas reserving different fashions for duties requiring greater response high quality or extra advanced reasoning.
Routing additionally provides operational necessities. Microsoft’s Azure Structure Middle notes that dynamic mannequin choice can complicate price forecasting, debugging, and efficiency evaluation when completely different requests are dealt with by completely different fashions.
Stripe and OpenRouter had been already working collectively earlier than the acquisition. In January 2026, Stripe mentioned builders utilizing OpenRouter might route mannequin requests by means of the platform whereas Stripe tracked utilization, utilized pricing, and dealt with billing.
The association paired OpenRouter’s routing layer with Stripe’s utilization measurement and billing programs earlier than the acquisition settlement.
Token utilization meets billing
Stripe has additionally been growing token-based billing instruments for AI purposes. Its LLM token-billing service, which Stripe at present lists as being in personal preview, can meter consumption in accordance with mannequin and token kind, together with enter, output, and cached tokens the place supported.
Stripe’s documentation says companies can use the system for per-token pricing, pay as you go credit, fastened charges with included utilization, or mixtures of these approaches. The corporate also can replace supported mannequin costs when suppliers change their underlying pricing.
OpenRouter already produces a lot of the utilization information concerned in these billing calculations. Its API stories immediate, completion, reasoning, and cached token counts with particular person responses, together with the price of the request.
It additionally data the underlying inference price charged by the supplier individually from the quantity charged to an OpenRouter account. Token counts are calculated utilizing every mannequin’s native tokeniser quite than making use of a single counting technique throughout all fashions.
Stripe CEO Patrick Collison has linked the acquisition to the position of tokens and computing assets in AI purposes. Collison mentioned tokens are a central unit for corporations constructing with AI and tied their financial use to how corporations handle obtainable computing assets.
Enterprise token consumption is already reaching massive volumes. Deloitte surveyed 515 US-based enterprise and expertise decision-makers in late 2025, all from organisations producing at the least $500 million in annual income.
The survey discovered that 37% of respondents had been consuming between one billion and 10 billion AI tokens per thirty days, whereas one other 30% had been consuming greater than 10 billion. By 2028, 61% count on month-to-month consumption to exceed 10 billion tokens.
Deloitte mentioned a lot of the rise would come from workloads exceeding 100 billion tokens per thirty days, with token use in that vary anticipated to triple from 2026 to 2028.
The agency cautioned that greater token consumption doesn’t essentially point out more practical AI adoption. Deloitte recognized outsized prompts, weak context administration, and restricted reuse as components that may improve token consumption.
The price of processing these tokens also can differ by mannequin and, in some instances, by the supplier serving the identical mannequin.
OpenRouter says it processes greater than 10 trillion tokens per day throughout a group of greater than 10 million builders and corporations. Earlier figures revealed by OpenRouter present some indication of how its visitors had modified earlier than the acquisition.
In Could, the corporate mentioned its weekly quantity had elevated from 5 trillion to 25 trillion tokens over the earlier six months. On the time, OpenRouter reported serving greater than eight million builders throughout greater than 400 fashions.
OpenRouter was based in 2023 and has raised funding from traders together with Menlo Ventures and Andreessen Horowitz. Its $113 million Sequence B spherical in Could was led by CapitalG, Alphabet’s impartial development fund, with participation from traders together with NVentures, ServiceNow Ventures, MongoDB Ventures, Snowflake Ventures, and Databricks Ventures.
Stripe and OpenRouter didn’t disclose the monetary phrases of the acquisition. Reuters reported that the transaction is price barely greater than $8 billion, citing an individual conversant in the matter who requested anonymity as a result of the data was confidential.
(Picture by appshunter.io)
See additionally: OpenAI president urges enterprises to hasten AI safety defences
Need to study extra about AI and large information from business leaders? Take a look at AI & Big Data Expo happening in Amsterdam, California, and London. The excellent occasion is a part of TechEx and is co-located with different main expertise occasions together with the Cyber Security & Cloud Expo. Click on here for extra info.
AI Information is powered by TechForge Media. Discover different upcoming enterprise expertise occasions and webinars here.
