Inside the Token Relay Market Powering AI Fraud and Resale

Overview

The relay market enables developers to obtain frontier‑model tokens at a fraction of official pricing by chaining together virtual‑card merchants, account pools, and relay gateways that resell access.

How the Relay Market Works

The ecosystem consists of four layers: upstream card and account merchants that supply virtual credit cards and bulk‑registered accounts; midstream account pools that aggregate those credentials, manage tokens and rate limits, and expose a single API; downstream relays (transfer stations) that wrap the pool’s API in a Chinese‑language product, handle billing and support, and compete on price; and end users—developers, startups, SaaS, and commercial buyers—who purchase cheap inference or use the traffic for model distillation.

Software Behind the Relays

Almost every relay runs on one of two open‑source OpenAI‑compatible gateways: one‑api or its more actively developed fork new‑api. Both expose a single endpoint that matches the OpenAI API, pull a key from a pool on each request, forward it upstream, and deduct quota using a multiplier. Across the tracked relays, one‑api appears roughly four times as often as new‑api, even though new‑api includes self‑service payment and multimedia model support. The tools themselves are neutral; a relay becomes illicit when its channels are stocked with stolen, leaked, or pooled keys and it resells that access against providers’ terms.

Methods of Abuse

Operators combine several tactics to keep costs low: automating free‑trial abuse to claim credits, executing charge‑back attacks or using stolen cards, funding accounts with prepaid cards, proxying traffic through unguarded support chatbots, and occasionally launching denial‑of‑wallet floods that purely burn provider spend.

Who Buys and Why

Buyers fall into three main categories: those seeking cheap tokens for inference, those trying to bypass geo‑restrictions on model access, and those using the cheap relayed frontier‑model outputs to train domestic models via distillation. Forum posts claim that some distillers have earned millions from this pipeline.

Market Maturity and Indicators

The market is already mature: price‑comparison sites exist, affiliate programs run, and gateway products are offered. The ten highest‑traffic relays tracked receive a combined 3.6 million visits per month. One relay directory, hvoy.ai, runs a daily provably fair lottery that gives away fifty $100 API keys each round; on a recent observed round, 258 participants entered 401 tickets for the fifty keys.

Defensive Measures for Providers

Providers can raise the cost of entry by making bulk account creation difficult and capping fresh‑account spend, watch for prepaid or virtual cards and mismatched billing info, monitor behavioral anomalies such as rapid registration‑to‑first‑token times, cluster accounts by shared IP or device fingerprints, and set up spend alerts and concurrency limits. When abuse is detected, quiet throttling is preferable to clear errors that reveal which signal to fix.

Community Perspectives

"I spent years working on financial integrity at a large ads company and this isn’t novel at all! The same resale markets are at play for the last generation of internet giant’s products." — @wtobey1

"One aspect that seems to be missing is the abuse of the free credits provided for new companies by AWS, Azure, and other providers." — @namanyayg

"The real problem is subscription models. Businesses want recurring revenue so they try to game the ratio of fixed subscription prices to COGS..." — @benlivengood

"It’s the same fundamental problem as “ticket touting” for popular events - if you sell something that’s in demand at a price that’s far lower than the clearing price of the market, you’re creating a juicy arbitrage opportunity..." — @ifwinterco

"This is the problem we’ve been working on solving with WorkOS Radar..." — @grinich

"the way i see it there are 3 types of resellers. the ones using fake credit cards to rack up costs and then cancel the card are doing actual fraud..." — @tancop

"Seems to be an LLM megaexpansion of the actual source (in chinese): https://www.v2ex.com/t/1196011" — @nojs

"How do the users know they’re getting what they’re paying for? I disabled automatic downgrading/rerouting because it sometimes takes me a second to tell when the answer came from a different model than I wanted." — @blfr

"For example, one operator’s price‑comparison site listed a package that bought the equivalent of $3,333 worth of official Anthropic credit for 425 RMB — roughly $0.13 of usage per $1 spent. Do these numbers make sense?" — @robluxus

"Ah I was wondering how a certain chinese site I came across did this and it could be this. That site offers substantial free tokens, is often reported as being flaky and their affiliate links are popping up on different social medias but look sketchy as anything." — @hahahaa

"This is more concerning to me from the perspective of being able to appear as “multiple entities” to the frontier models. My question is do the companies know and are able to detect and consolidate all these accounts as a single actor and just don’t care to combat it?" — @chrismarlow9

"This is pretty fascinating! Here are the two open source proxies listed in the article: https://github.com/songquanpeng/one-api and https://github.com/QuantumNous/new-api" — @simonw

"Nice research and structuring into 4-tier layer. For providers like Anthropic and OpenAI, subscription is the entry point for all these, right? Besides the measures proposed in the article, can token usage % determine these clusters of accounts?" — @bg24

"Don’t the operators store the agent traces and resell them as training data to AI companies? Why wasn’t this mentioned? Did you find any evidence of that?" — @1337h4xx

"thats one of the reasons why we vest any of our new customers. We need to know you before you are allowed to use our agent system. When you have an open sign up with some free credits, all hell breaks loose." — @lmf4lol

"This article is about the mechanics, but the title implies this is unethical. Why is this practice considered unethical?" — @cobzilla

"Aren’t these figures the wrong way around “$0.13 of usage per $1 spent” So I spend a dollar and I get 13 cents worth of usage?" — @rustyhancock

"How do you know the reseller is even giving you the genuine article? Could they be advertising Fable but repacking Deepseek?" — @3eb7988a1663

"How do we define “fraud” if taking a new user discount over and over is fraud?" — @inigyou

"I use both of subscription and API services. on last month, i chat with CLI and let it to do something. After that, maybe in one days pass, i received the $32 USD bill. it cause my left my API key and CLI call the API to do job not through subscription." — @hsienchuc

"The chinese AI and greymarket peptides scenes feel so vital, fun and wild-west. Imagine spending your life being beholden to USA corporations and being scolded on HN for violating a trillion dollar company’s ToS." — @chhxdjsj

"The relays sound also like a nice source of monitoring for whoever controls them." — @boznz

"If we compare OpenAI subscription prices vs. the cheapest inference providers on OpenRouter, than OpenAI must be losing money. Add the fraud to this and the question is what the future will hold. The only salvation is hardware getting 10x cheaper before the labs run out of money." — @edg5000