1 Introduction
The AI learning map on my blog hasn't been updated for a while. It's not because my interest in AI has waned; on the contrary, many ideas and explorations have already taken shape in my mind. The reason that actually stopped me in my tracks is very simple:Many practical scenarios ultimately cannot bypass the barrier of “paid APIs”。
Although there are many free API services on the market, they either have extremely low call limits, outdated models, tedious deployment, or even fragmented experiences, which really makes me lose interest in continuing to “tinker.” And when it comes to “paying,” of course I don't mind,Unfortunately, many service providers instead exclude us domestic users. Especially when it comes to the payment process, there are various risk controls, restrictions, credit card verification failures, and even the API Key cannot be applied for.
Recently, I finally figured out the WordPress multi-active deployment (see details:Home Data Center Series: WordPress Active-Active Architecture (Simplified Version) Implementation Plan in Personal Blog), and having some free time, I decided to return to the theme of AI exploration. As the first step, I want to solve the most basic and troublesome problem:How to stably and with a low barrier use the paid API services of mainstream foreign AIs。
2 Why I ultimately didn't choose OpenAI, but chose OpenRouter instead?
1 Why I chose not to use the official OpenAI API?
The GPT series holds a benchmark position in the AI field, but as a domestic developer/user,wanting to directly use these models through the official OpenAI API is fraught with difficulties, and the core problem actually boils down to just one word:Payment (of course, there are also regional access restrictions, but this can be solved through scientific methods or magic, so it is not considered a core problem)。
OpenAI officially does not support Alipay or WeChat, and even domestic-issued credit cards (including dual-currency cards) cannot be successfully bound or used for payment. Consequently, many domestic developers have had to resort to “indirect routes”:
- Either try to get an overseas credit card (time-consuming, high barrier, lots of risk control)
- Or rely on various “magic payment” channels (such as virtual cards, top-up codes, third-party payment agents, etc.)
- Even worse, some services that were previously usable (such as WildCard) have now gradually “gone under”
These methods seem to solve the urgent need, but in essence, none of them are long-term solutions:They are neither stable nor compliant, and cannot support the long-term use of a formal project。
2 The leader among multi-model access platforms: OpenRouter
After some research, my conclusion is:Instead of struggling with OpenAI's official channels, it's better to choose a reliable third-party compatible platform.. And among these platforms,OpenRouter is currently one of the most highly recommended choices.。
OpenRouter.ai is a technology startup headquartered in New York, USA. Founded in 2023, it is dedicated to building a unified multi-model access platform. It does not develop large language models itself, but rather integrates the official APIs of mainstream model providers such as OpenAI, Anthropic, Google DeepMind, Meta, and Mistral to provide developers and application systems with a “neutral middle-end” calling method. With OpenRouter, users can flexibly switch between different models using a unified API syntax, lowering the barrier to entry and avoiding the complex process of dealing directly with multiple vendors.
The company was co-founded by OpenSea co-founder Alex Atallah and engineer Louis Vichy, backed by top Silicon Valley venture capital firms such as a16z (Andreessen Horowitz), Menlo Ventures, and Sequoia Capital, and completed a $40 million Series A funding round in mid-2024. The company operates on a fully remote model, with core members from across the United States and Europe, maintaining the agile pace of a startup team.
At the product level, OpenRouter's advantages lie not only in its wide selection of models and low barrier to entry, but also in its extreme friendliness to the developer ecosystem—in addition to standardized API access, it also provides detailed documentation, flexible request pricing, and real-time model routing. This is particularly important for developers who want to integrate AI features into practical applications, especially in the domestic environment, where directly calling official APIs is often fraught with obstacles due to payment, network, and risk control issues.
It is worth mentioning that OpenRouter does not “proxy” model data, but strictly completes request inference through officially authorized APIs, and data is not intercepted or stored by the platform; it is merely a “distributor” of model requests and does not participate in underlying computation or response modification. This model also makes it more convincing in terms of stability and security. In other words,the GPT, Claude, Gemini, and other models you call are essentially still powered by their respective official systems for inference services, and OpenRouter is just the “routing layer” in the middle rather than a “disguise layer”. Compared to some “informal channels” that rely on reverse engineering or crawler simulation, OpenRouter is closer to an “official transfer” mechanism, with no need to worry about common risks such as data leakage, high model latency, or API failure, making it more suitable for production deployment in formal scenarios.
Because of this positioning, OpenRouter has rapidly attracted millions of developers over the past year, becoming one of the most active and mature AI multi-model access platforms today. For domestic users, its most attractive aspects are: in terms of payment, it is compatible with dual-currency cards (such as Visa/Mastercard) issued by domestic banks, and also supports Alipay/WeChat top-ups; in terms of access, the domestic network can connect directly (though direct connection is not recommended, for reasons explained later), almost eliminating those headache-inducing “non-technical obstacles” in traditional API access.
OpenRouter is not the only multi-model access platform on the market that supports Alipay/WeChat and direct domestic connection—for example, OhMyGPT, which is familiar to domestic users (and is also a third-party multi-model API provider I previously recommended; friends who are not familiar with OhMyGPT can refer to my previous article:Home Data Center Series: Starting Your AI Journey: A Detailed Introduction to Local Large Language Model UIs and Large Language Model API Providers), which also supports Alipay/WeChat payment and direct domestic network connection—but I still highly recommend OpenRouter for 4 reasons.
1、Credibility. OpenRouter is backed by well-known entrepreneurs and top Silicon Valley venture capital firms, and has established long-term, official API-based partnerships with vendors like OpenAI, Anthropic, and Google DeepMind. This means its API call process is transparent, data will not be privately intercepted, and service stability is better guaranteed.
2、Industry Recognition. It is not only highly praised by many developer communities, but also natively supported by a large number of third-party AI tools, such as the well-known WordPress plugin AI Engine which directly built in OpenRouter's access option, which is rare among similar platforms:

This ecosystem advantage makes it easier for you to switch between different applications without having to maintain multiple sets of configurations yourself, which is also the reason why I eventually switched from OhMyGPT to OpenRouter.
3、Balance between Globalization and Localization. It maintains the standards and security of an international platform while optimizing the payment and network experience for domestic users. Compared to some “grey area solutions” that rely on reverse engineering, its sustainability is higher, and it won't easily fail or restrict features.
4、Transparent Pricing and Low Premium. OpenRouter's token pricing is basically consistent with OpenAI's official pricing, only charging about a 5.5% transaction fee during top-up; whereas platforms like OhMyGPT usually add about 10% on top of the official price for cost and profit. For high-frequency calls or large model scenarios, this difference accumulates to a considerable amount.
All in all, OpenRouter is a “hassle-free and reassuring” choice for me, which is why I ultimately decided on it as the recommended multi-model access platform after my research.
3 Usability comparison of mainstream AI API platforms for domestic users (2025 Edition)
To illustrate the advantages of OpenRouter more intuitively, I have compiled a comparison table of common AI API service providers, which includes both multi-model access platforms and official model APIs. It covers several key factors such as registration difficulty, connectivity, payment methods, and recommendation level:
| Name | Registrable | Direct Connection Available Domestically | Supported Payment Methods | Recommendation Level | Remarks |
|---|---|---|---|---|---|
| OpenRouter | ✅ | ✅ | Dual-currency credit card (recommended), Alipay/WeChat | ⭐⭐⭐⭐⭐ | Easy access, supports multiple models, high ecosystem recognition, the first choice for domestic developers |
| API2D | ✅ | ✅ | Chinese bank cards, Alipay, WeChat and other third-party channels, Stripe | ⭐⭐⭐⭐ | Close to the native OpenAI API, price is slightly higher than OpenRouter |
| OhMyGPT | ✅ | ✅ | Alipay, WeChat, domestic bank cards | ⭐⭐⭐⭐ | Targeted at domestic users, transparent pricing, but the number of models and update speed are inferior to OpenRouter, and the ecosystem influence is weaker |
| Hugging Face | ✅ | ✅ | Overseas foreign currency credit cards | ⭐⭐ | Limited free quota, requires self-deployment of models, suitable for users familiar with the deployment process |
| Claude (Native) | ❌ | ❌ | Limited to overseas credit cards and addresses (difficult to register domestically) | ⭐⭐ | Recommended to access indirectly through OpenRouter |
| Gemini (Google) | ❌ | ❌ | Overseas Google Pay, requires an overseas credit card | ⭐⭐ | Official APIs are fragmented, inconsistent experience, and basically cannot be used via direct connection domestically |
| Azure OpenAI | ❌ | ✅ | Corporate accounts linked to foreign currency cards, corporate invoices | ❌ | Enterprise-oriented, tedious integration, requires approval process |
| OpenAI Official | ✅ | ❌ | Only foreign currency credit cards issued overseas(Credit cards from domestic issuing banks are basically unusable) | ❌ | Difficult registration, difficult card binding, many API restrictions, direct use is not recommended |
In addition to the GPT series, it also integrates popular models such as Claude, Gemini, Mistral, Command R, and LLaMA,Unified interface standards, pay-as-you-go, complete documentation, no special payment methods required, making it highly suitable for domestic users for quick integration, development testing, or product integration.
More importantly, it not only supports direct domestic access, but is also extremely friendly to domestic users in terms of payment methods:
- ✅ Supports dual-currency credit cards (Visa/Master)
- ✅ Supports Alipay and WeChat scan-to-pay top-ups (Provided officially by OpenRouter, but not directly displayed on the checkout page; it needs to be triggered through a specific entry point. This method has a stable process and a low barrier to entry, making it very suitable for domestic users. For details, see the “Credits Page in Detail” section below)
OpenRouter provides domestic users with a truly “out-of-the-box” experience: No registration hassle, no card binding, no risk control, multi-model support, pay-as-you-go, and top up as you use. Compared to those paths of repeatedly “struggling with payment methods,” it is undoubtedly the most worry-free and suitable solution for practical development at present.
Updated on 2026-07-08
It's been a long time since I logged into the OpenRouter interface. I logged in today and found some bad news. Presumably in response to requests from OpenAI, Anthropic, and Google, OpenRouter has now also started to judge eligibility to call models from these three companies based on the region of the billing address:

Therefore, for the models of these 3 companies, they can no longer be called through OpenRouter in the future unless you have a qualifying billing address. For friends who need to use models from these 3 companies, it is still recommended to call them through OhMyGPT in the future.
3 OpenRouter Account Registration
Registering an OpenRouter account is very simple. On the official website (https://openrouter.ai/), just follow the graphic process below to register by yourself (currently, direct access is possible from within China):



I have a Gmail account, so I just click the Google icon to proceed:



In this way, the OpenRouter account is successfully registered:

After that, you need to set up various commonly used parameter items for the account.
Generally speaking, accessing the official OpenRouter website directly using domestic broadband for registration is not a problem and does not require a special network environment. However, if you access it while using a VPN (”scientific internet access”), and your DNS settings use the Fake-ip mode, plus you are using the Chrome browser, you might suddenly encounter a cold prompt: ”Sorry, you have been blocked.“:

This is very mysterious—it's not that the IP is blocked, nor is there an issue with the account, but rather some underlying mechanisms combined together triggered Cloudflare's risk control strategy (OpenRouter's website is behind Cloudflare).
Simply put, this is because in Fake-ip mode, the target IP accessed by the browser is actually forged, and Chrome actively tries to perform connection verification, which ends up being misidentified by Cloudflare as a “suspicious request” or even “automated behavior,” resulting in an immediate block.
The solution is not complicated; you can:
- Temporarily switch to Firefox or Safari browser to try;
- Or more thoroughly, set openrouter.ai to direct connection (direct) in Clash to avoid DNS interference;
- If you are using an OpenWRT router based on Clash, you can also create a bypass rule for this domain.
This issue is actually quite interesting, involving the working principle of Fake-ip, Chrome's DNS behavior, and Cloudflare's blocking judgment mechanism. I will write a dedicated article later to analyze this phenomenon in detail.
4 Introduction to OpenRouter Account Settings
1 Overview
After registering an OpenRouter account, the next step is to make some necessary settings for the account. OpenRouter's backend settings panel is not complicated and is divided into the following 7 parts:
- Settings: The basic settings page, where you can modify your nickname, avatar, default language model, etc., suitable for priority configuration right after registration;
- Credits: The place to view balance and consumption records, and also the top-up entry point, including the function to purchase credits through domestic channels (such as Alipay);
- Presets: Preset configurations, supporting the definition of custom parameters (temperature, context length, etc.) for different models, which can be easily referenced when calling the API later;
- API Keys: Used to generate and manage API keys, which are the core credentials for calling OpenRouter;
- Provisioning Keys: Mainly used for team collaboration or SDK integration scenarios, allowing fine-grained control over permissions;
- Integrations (BYOK): Supports the “Bring Your Own Key” (BYOK) mode, suitable for users who already have model API keys to connect them to OpenRouter's unified interface;
- Privacy: Privacy options, where users can choose whether to allow OpenRouter to record request logs and other behaviors.
Each of these sections has its own focus in practical use: if you just want to simply call models via the API, you only need to configure the API Keys to get started; but if you want to finely manage different projects, top up using Alipay, or manage multiple model parameters with presets, the other sections will also be very useful.
Starting from the next section, we will introduce these functional blocks one by one, beginning with the basic and important account settings page:Settings。
2 Settings Page Details
✅ Account (Account Management)

Click this section to manage your account information, such as:
- Modifying login email / password;
- Setting or changing security verification methods (such as enabling 2FA);
- Completely canceling and deleting the account.


Although you won't change these often in daily use, it is recommended to set up security options at the beginning to prevent account abuse.
✅ Organization (Organization Management)

OpenRouter supports creating “Organizations” to facilitate team collaboration, unified billing, and other scenarios. For individual users, this can be ignored; but if you plan to have multiple members share an API Key or use unified billing in the future, you can create an organization here and invite other accounts to join.
✅ Low Balance Notifications

When the account balance is low, the system will automatically send an email notification. By default, it will be sent to your registered email. It is recommended to keep this feature enabled to prevent API call failures without your knowledge.
✅ Allowed Providers (Allowed Model Providers)

OpenRouter is connected to many model providers, such as OpenAI, Anthropic, Mistral, Cohere, etc. If you wish to use only some of them, you can manually select the allowed providers here.
In addition, there is an “Always enforce” option:
- Once enabled,it will only use the providers you have checked here;;
- when not enabled, it defaults to using all available providers that are not ignored.
This feature is suitable for those who want precise control over where requests go (for example, only wanting to use OpenAI and not any open-source models).
✅ Ignored Providers

Opposite to the above, here you can select the model providers you do not wish to use. For example, if you do not trust certain niche providers or feel that a certain provider's response is too slow, you can exclude them here.
✅ Default Provider Sort (Default Provider Sorting Logic)

OpenRouter will automatically choose a model provider for you based on certain rules when you do not specify one. This “default sorting logic” can be controlled here, and the default value is:
- 默认(均衡): Automatically balances lower prices with service stability, suitable for most users.
In addition to the default option, you can also choose the following sorting strategies based on your needs:
- 价格(低价优先): Prioritizes the lowest-priced providers, suitable for cost-sensitive users;
- 吞吐量(最高优先): Prioritizes providers with the highest throughput (concurrency handling capability), suitable for application scenarios with high concurrent request volumes;
- 延迟(最低优先): Prioritizes providers with the lowest response latency, suitable for real-time interaction scenarios with high response speed requirements.
Generally speaking, if you have no special preference, you can just keep the default; but if you are pursuing the ultimate cost-effectiveness or are extremely sensitive to response speed, you can also flexibly switch the sorting logic according to your needs.
✅ Default Model

This is a relatively important setting. You can specify a default model here, and in the future, when you call the OpenRouter API, if you do not manually specify a model, the one set here will be used automatically.
Clicking ”点击这里“ opens the model list, where you can see which models are supported and how much each model costs.
⚠️ Tip:
- If you are using applications that are highly sensitive to model selection (such as auto-summarization, translation bots, etc.), it is recommended to explicitly set a default model;
- If you frequently switch models, you can also leave it unset and specify it manually with each request.
Summary
Although the Settings page is generally geared towards “basic configuration,” several key options will profoundly affect your behavior when calling OpenRouter later, especially the “Default Model” and “Provider Selection” sections.
It is recommended to come here right after registering your account to:
- Set up your most commonly used default model;
- Allow or ignore specific providers based on your usage preferences;
- Enable low balance notifications to avoid API request failures.
In the next section, we will enter the Credits pageto learn how to view your balance, top up, and make payments through domestic channels (such as Alipay).
3 Credits Page Details
3.1 Overview
Credits pageis the main entry point for OpenRouter users to manage account balances, top-ups, and consumption. Simply put, it is your... when using various model APIs on OpenRouter.Credits Center, where almost all billing-related operations can be completed:

The overall interface is very clean and intuitive, mainly including the following parts:
- Account Balance (Available Credits): Displays your currently available balance in USD. The costs incurred when calling models will be deducted from here in real-time.
- Top-up Portal (Buy Credits): Click the ”Add Credits” button to jump to the top-up interface. It supports foreign currency credit cards, some RMB cards, and top-ups via Alipay/WeChat redemption codes (to be detailed in Section 5.3.2).
- Auto Top-Up: Allows you to set an auto top-up rule to automatically charge and renew when the balance falls below a specified threshold, avoiding API call interruptions due to exhausted balance (to be detailed in Section 5.3.3).
- Payment History: Displays all your past payment records, including time, amount, and payment method, making it easy to check accounts.
- Usage Details (View Usage): Allows you to view the billing details of each API request by dimensions such as model, time, and project, helping you analyze and control costs.
Overall,The Credits page has clear functions and simple operations, making it easy for even first-time OpenRouter users to get started quickly. You can think of it as a “transparent billing dashboard” that provides a complete view of cash flow for your model calls.
3.2 Buy Credits (Top-up Portal)
If you do not want to enable the auto top-up function,Buy Credits is your main entry point—used to manually top up your account balance and flexibly control your capital investment as needed.
OpenRouter provides a variety of top-up methods, covering two major categories: traditional payment channels and cryptocurrency, suitable for developers in different regions and with different preferences.
3.2.1 Traditional Payment Methods (Recommended)
After clicking the “Add Credits” button, you will see the following four mainstream top-up methods:
- Card

The most common and recommended method, supporting the binding of Visa / MasterCarddual-currency credit cards. Most overseas credit cards can complete payments smoothly, including some issued by domestic banksdual-currency cards(whether they are usable depends on the issuing bank's risk control policy).
**Item 9:**RMB single-currency cards can hardly complete payments, so if domestic users choose this method, it is highly recommended to use acredit card with foreign currency payment capabilities。
- Bank

Payment via US local bank account (ACH), suitable for US users,not suitable for domestic users。 -
Cash App Pay

A payment app for US users, requiring a US phone number and bank account to be bound,not available for domestic users。
- Amazon Pay

Requires binding an Amazon account and an international credit card, with limited functional coverage,and a low success rate for domestic use。
Overall,for domestic users, the only practical method is Card (foreign currency credit card), and it is recommended to prioritize this path for binding and payment.
3.2.2 Cryptocurrency Payment (Use crypto)
If you cannot use a credit card, you can also choose to top up your OpenRouter account viacryptocurrency. On the top-up page, click the “Use crypto” switch to open the crypto payment channel, using mainstream crypto assets including BTC, ETH, USDT to complete the transfer:

Unlike traditional payment methods, OpenRouter's cryptocurrency payment process is supported by third-party platforms (such as Coinbase), offering a smooth overall experience and flexible access methods. However, it should be noted that this method willcharge an additional 5% transaction fee. The official instructions state that the actual payment amount is calculated as:
total_cost = credits_requested / (1 - 0.05)
In other words, if you want to top up $100 in Credits, you will actually need to pay approximately $105 in cryptocurrency.

After entering the payment process, OpenRouter provides a variety of wallet options for users to choose from, including:
- Coinbase: Suitable for users who already have a Coinbase account, allowing direct redirection to pay;
- Coinbase Wallet: Used to connect Coinbase's own crypto wallet;
- Smart Wallet: A built-in wallet that does not rely on browser extensions, suitable for users unfamiliar with Web3 operations;
- MetaMask: The most common browser extension wallet, recommended for most desktop users.
In addition, you can also click “Connect another wallet” to connect more types of wallets:

It supports multiple mainstream wallets including Rainbow, WalletConnect, Browser Wallet, Argent, Ledger (hardware wallet), Phantom, Rabby, Trust Wallet, Uniswap Wallet, and more. Whether you hold tokens on Ethereum, Polygon, Solana, or other mainstream chains, you can basically find a suitable payment path.
If you are already familiar with Web3 wallet operations, or hold stablecoin assets (such as USDT/USDC) yourself, crypto payment is undoubtedly a secure, fast, and KYC-free method. However, for most domestic developers, if they have no experience using cryptocurrency, it is still recommended to prioritize dual-currency credit card payments, which are simpler to operate and carry lower risks.
3.2.3 Alipay / WeChat QR Code (Relay Purchase)
In addition to the officially supported card payments and cryptocurrency methods, OpenRouter also provides arelatively hidden but very friendly “third-party payment” channel for domestic users, through which you can use Alipay / WeChat Pay QR code paymentto directly complete the top-up.
This featureis hidden by default, requiring you to follow these steps on the Credits page:
- OpenOpenRouter's Credits page;
- Click the Add Credits button;
- At the bottom of the pop-up top-up window, click Use one-time payment methods

- The page will expand a new area, providing options for Alipay / WeChat Pay QR code payment

- Follow the instructions to go to the external payment platform (Alipay or WeChat Pay) to complete the purchase;


⚠️ Notes
- The payment process will redirect to thethird-party partner platformintegrated by OpenRouter, such as the payment page for Alipay/WeChat Pay QR codes; this entry is currently accessible through the official UI page (Add Credits → Use one-time payment methods), so it can be regarded as a platform-approved transit method;
- After the payment is completed, the topped-up amount willbe credited directlyto your OpenRouter account balance, usually with a delay of no more than 1 minute;
- The service fee is already included in the total price(for example, if you pay 78 RMB and receive $10, the implicit exchange rate is about 7.8), with no extra charge, though the exchange rate will differ from the mid-market rate;
- Although it is not a native OpenRouter payment channel, since the entry comes from the official website interface,this method has been widely verified by the community as stable and secure, making it suitable for beginners without foreign currency credit cards or users making small top-ups.
Recommended Audience
- Unable to use foreign currency credit cards;
- Unfamiliar with or inconvenient to use cryptocurrency;
- Hope to complete the Top-Up through familiar domestic payment methods (Alipay / WeChat).
Overall, this transit methodis simple to operate, fast to arrive, and requires no registration of an additional wallet, making it one of the more practical solutions for domestic users at present. However, as it involves third-party payment, please always remain vigilant and avoid making payments through non-official website portals.
3.2.4 Summary
Overall, although OpenRouter lists various Top-Up channels on its interface, there are actually only three options truly suitable for domestic users, and their applicable scenarios are different.
First priority: Foreign currency credit card (Card)
If you own a credit card that can make foreign currency payments (such as a dual-currency, multi-currency, or all-currency card from Visa or MasterCard),it is highly recommended to prioritize using the Card method for Top-Up. This is currently the most official, direct, and cost-effective Top-Up channel:
- Convenient Top-Up process, instant arrival;
- No transit or third-party tools required;
- No extra transaction fees, and it is also the only “zero-cost” Top-Up method;
- Can be bound for auto-renewal, and is friendly to API risk control.
Second priority: Alipay / WeChat QR code Top-Up (transit channel)
If you do not have a foreign currency credit card but still wish to complete the Top-Up through RMB, you can also choose to use Alipay or WeChat QR code payment. This method is implemented through the transit channel embedded in the official OpenRouter page,with the entry located below the “Use one-time payment methods” button on the Add Credits interface, offering a simple operation process and a good experience.
Although it belongs to a platform-partnered transit payment channel (not a native official settlement channel), since the entry comes from the official website interface itself,it can be regarded as a controlled and relatively safe officially recommended method. Once the payment is completed, the system will automatically top up the balance to your account,with no need to manually paste a Top-Up code, which is suitable for novice users who do not meet credit card conditions, especially for small-amount trial scenarios.
- No credit card or cryptocurrency required;
- No transaction fees, fast arrival speed (usually within a few minutes);
- Suitable for users without foreign currency payment capabilities.
Third priority: Cryptocurrency payment (Use crypto)
For users who possess certain crypto assets and are familiar with Web3 wallet operations, they can also choose to top up through currencies such as USDT, ETH, and BTC. This method does not have a high threshold, but requires bearinga platform transaction fee of 5%, and the arrival speed is usually fast. Suitable for technical users or special demand scenarios.
- No real-name verification or traditional banking system required;
- High transaction fees, not recommended for frequent small-amount use;
- Supports multiple mainstream wallets, including MetaMask, WalletConnect, Coinbase, etc.
As for those listed in the interface methods such as Bank, Amazon Pay, Cash App Pay, etc.,are basically unusable for domestic users. These options mostly require a local US account or payment environment, lack universality, and can be directly ignored.
Conclusion: Foreign currency credit card first, QR code Top-Up as the most convenient alternative, crypto payment reserved for experienced users, and other methods can be ignored.
3.3 Auto Top-Up (Automatic Top-up)
In addition to manual on-demand Top-Up, OpenRouter also provides a relatively worry-free option:Auto Top-Up (Automatic Top-Up). This feature can automatically deduct funds from the bound payment method when the account balance falls below a certain threshold, replenishing credits for you in a timely manner, thereby avoiding interruption of the model API use due to exhausted balance.
This feature is particularly suitable for developers with relatively stable usage needs, such as integration into production environments, regularly running tasks, or scenarios with high call frequencies, effectively reducing the risk of “service interruption due to forgetting to top up”.
It should be noted that,Auto Top-Up only supports payment methods with automatic deduction capabilities, such as credit cards (Card). One-time payment channels like cryptocurrency or QR code payments cannot enable this feature.
The entry for Auto Top-Up can be directly seen on the Credits page:


The specific method for adding payment methods has been explained in the previous section and will not be repeated here. As long as you have bound a supported payment method, you can set the auto Top-Up amount and trigger conditions.
Note: Although auto Top-Up is very convenient, it also means that the platform can initiate deductions without your active operation. Therefore, it is recommended to ensure your card security and sufficient limit before enabling it, to avoid unnecessary financial losses caused by misoperation or abnormal balances.
4 Presets Page Details
4.1 Overview
In the daily use of OpenRouter, we often want to set different usage preferences for different tasks, such as fixing a certain model, adding custom system prompts, or adjusting generation parameters and controlling budgets. To reduce the hassle of repeated configuration each time, OpenRouter provides the “Presets” feature, allowing users to save these settings as a preset template to call at any time.
The Presets page is not complex, but the features it provides are very powerful. Here, you can name each Preset, add a description, set the default model or provider to use, and control advanced parameters such as routing strategies, privacy options, and price caps. At the same time, Presets also support setting specific parameters during generation, such as temperature, maximum tokens, and repetition penalty, making it suitable for users with specific requirements for output style or result quality.
In short, Presets are the “basic settings” you lay down for how you use the models. Once configured, they can be called with one click in the future, without needing to re-enter them every time. For those with fixed usage needs, it will greatly improve efficiency.
Important Reminder:
Presets is an API-level feature, which only takes effect when the Preset ID is passed during the API call. If the third-party tools you use (such as some WordPress plugins, LobeChat, Chatbox, etc.) do not provide a Preset parameter input themselves, these presets will not be automatically applied during calls; instead, they will execute according to the parameters set within the tool (generally, these tools set the specific model and other information internally, so the settings in Presets are ineffective for these tools). Therefore, Presets are suitable for use when writing your own code, using clients that support custom API parameters, or debugging interfaces, which can significantly reduce the workload of repeated configuration.
4.2 Create a New Preset
On the Presets page, click “创建预设“ button to start creating a new usage preset:

Next, I will explain the function of each setting item one by one, and demonstrate how to fill them out based on the needs of a “Writing Assistant”.
Basic Info: Basic information of the preset
This is mainly to name the preset and write a simple description, making it easy for you to distinguish and manage.
- 名称: Write an easy-to-remember name, such as Writing Creative Assistant, Creative Writer, supporting Chinese.
- 别名: The unique identifier generated by the system, which defaults to the same as ”Name”. However, if ”Name” is in Chinese, it is recommended to manually change it to English here.
- 描述: Briefly explain the purpose, for example: “Suitable for Chinese content creation, outputting smoothly and naturally”.

System Prompt: System preset prompt words
This is aglobal instruction to control model behavior. You can understand it as a “default context” or “role definition”.
For the Writing Creative Assistant, you can set the following content:

This will significantly affect the language style and response perspective of the model's output.
Models: Specify supported models
If you only want this preset to be used under specific models, such as ChatGPT-4o or Claude 3 Opus, you can check them here.
For the “Writing Creative Assistant”, you can choose the following models:
- OpenAI:ChatGPT-4o: Strong balanced capability, suitable for creative tasks
- Anthropic:Claude 3 Opus: Excellent language organization skills, and also performs very well in Chinese

Of course, you can alsochoose not to select any model, meaning this Preset can be applied to all models.
Provider Routing: Model provider selection and routing strategy
This part belongs to the Preset'sadvanced options, which is disabled by default. You need to check ”包含服务商偏好” to see the detailed configuration items. You can understand it as: “telling the system which model providers I trust, and whether to allow automatic switching to backup providers if there is an issue.” Most users can completely skip this, as OpenRouter's default strategy is already smart enough to make reasonable choices among quality, speed, and price.
However, if you have the following demands, manual configuration is recommended:
- Only trust certain providers (such as OpenAI, Anthropic)
- Wish to exclude unstable providers
- Want to explicitly control the priority order of provider calls (whether this is effective needs testing)
- Sensitive to privacy, do not want data to be collected or used for training
The descriptions of each parameter are as follows:
| Parameter | Meaning and Recommended Settings |
|---|---|
| 允许回退 | Whether to allow automatic switching to backup providers; it is recommended to keep this checked. Otherwise, an error will be reported directly in case of a failure. |
| sort | Backup routing strategy, just keep the default |
| data_collection | Whether to allow data to be recorded and trained by providers: — allow (default) — deny (more private). Choosing deny may limit the types of models available. |
| only | Whitelist, only allow these providers to provide services |
| ignore | Blacklist, exclude providers you do not want to use |
| order | Specify call order —— ⚠️ In practice, sorting may be ineffective, be careful! |
In theory, the purpose of the order field should be to allow users to explicitly specify the call order of multiple model providers, such as which one to prioritize and which one to try after a failure. However, according to my actual testing, the current performance of this feature is inconsistent with expectations.
Specifically: after selecting multiple providers to add to the order list in the interface, users cannot manually drag to adjust their order. The system seems to determine the actual call priority based on the providers' namesalphabetical orderto determine the actual call priority. This means that even if you select OpenAI first and then Anthropic, Anthropic is highly likely to be prioritized—simply because it comes first in the alphabet.
This point is not currently explained clearly in the official documentation, so I cannot confirm whether this is a UI limitation, a design flaw, or if it is executed according to this logic by design.As a precaution, we should assume that the current order field cannot precisely control the call order.
Therefore, if you only want to use a specific provider, such as OpenAI, you should only check this provider in the only option to avoid mixing in other candidates. If you need to set up multiple backup providers, please do not rely on the order field to strictly control the call order. A better approach is to reasonably use only and ignore to define the scope of trusted providers, and keep allow_fallbacks enabled, so that when the preferred service is unavailable, the system can automatically switch to other available providers to continue executing the request.
Correct setting method (Recommended)
If you want to achieve “prioritize a certain provider, and automatically switch upon failure”, please operate as follows:
You only trust OpenAI 和 Anthropic, and wish to prioritize using OpenAI。
✅ Correct settings are as follows:
- Check ✅ Include Provider Preferences
- Enable ✅ allow_fallbacks
- Set order = [OpenAI]
- Set only = [OpenAI, Anthropic]
- Leave ignore blank


Parameters: Model generation parameters
These are detailed control options for model behavior:

The meanings of the parameters in the image above are explained below, and you can set them according to your actual needs:
| Parameter Name | Default Value | Description of Effect |
|---|---|---|
| Temperature | 1.00 | Controls therandomnessof the output. The higher the value, the more creative the generated content; the lower, the more deterministic. Recommended range is 0.2–1.2. |
| Top P | 1.00 | Also known as the “nucleus sampling” parameter. Controls thediversityof the generated content; the smaller the value, the more the output is concentrated on high-probability words. |
| Top K | 40 | Limits the maximum number of “high-probability words” considered during sampling. Smaller is more conservative, larger is more flexible. |
| Frequency Penalty | 0.00 | Penalizes words that appear frequently in the text to reduce repetition. Suitable for generating more varied text. |
| Presence Penalty | 0.00 | Penalizes words that have already appeared in the text, encouraging the model to explore new content and reducing repetition. |
| Repetition Penalty | 1.00 | Another way to control repetition. Greater than 1 suppresses repetition, while less than 1 may encourage repetition. |
| Max Tokens | 1024 | Controls the maximum number of tokens generated in this response (1 token ≈ 1 English word, or 1–3 Chinese characters). |
| Seed | Random | Used forgenerating reproducible responses. The same seed and prompt can generate the same content on certain models (requires model support). |
After completing the above settings, click the blue button in the top right corner “Save Preset” to complete the save. Next, during conversations, you can directly switch to this “Creative Writing Assistant” Preset to enjoy a more customized experience.
4.3 Summary
At this point, we have a general understanding of the basic settings for the model parameters (Parameters) section.
These parameters are mainly used tofine-tune the output style of the model, allowing you to more finely control dimensions such as the randomness, repetition, and diversity of the generated content. For average users, using the default values in most scenarios will already yield ideal output results, so there is no need to obsess over these technical details at the beginning. However, if you have clear requirements for the style, structure, or consistency of the generated content—such as a more stable writing style, less repetitive content, or reproducible outputs—then reasonably adjusting these parameters can be very helpful.
It is important to note that not every model supports all parameters. For example, some lightweight models may not support the Seed parameter, which means that even if you set a specific random seed, you cannot guarantee reproducible results. Similarly, different models may respond slightly differently to parameters like Temperature and Top P, so certain settings might have a noticeable effect on one model but be ineffective or show little difference on another.
Therefore, if you are not familiar with the functions of these parameters, it is recommended to keep the default settings. Observing and fine-tuning gradually through actual use is safer and easier to get the hang of than making extensive modifications right from the start.
5 API Keys Page Details
5.1 Create Your Own API Keys
Although the API Keys page is very simple in terms of functionality, it is a crucial part of the entire system — any program, plugin, or tool that calls a Preset or model service relies on an API Key for identity authentication and call authorization.
The process of creating an API Key is very simple:

Click the “Create API Key” button, and the system will pop up a small panel:

You only need to fill in the following fields:
- Give this Key an easily identifiable name (Name), such as “Creative Writing Assistant”;
- If needed, you can set a credit limit (Credit limit) in USD; leaving it blank means no limit is set;
- In the advanced settings, you can also decide whether to include the usage of BYOK (Bring Your Own Key) modelsin the credit limit of this Key (by checking Include BYOK usage in limit).
After clicking Create, the system will generate a secret key string,which will only be displayed once, so please save it properly in a timely manner.
The function of the ”Include BYOK usage in limit” option in the advanced settings is:Whether to also count model calls under the BYOK (Bring Your Own Key) mode toward the credit limit of this API Key.
Simply put:
- Checked: Whether it is a model through OpenRouter or a model Key you bound yourself,it will be included in the credit control of this API Key;
- Unchecked: Only models called through the OpenRouter platform will be counted toward the credit, and BYOK mode will not be counted.
If you want to control calling behavior more strictly (for example, to prevent external tools from abusing your bound OpenAI Key), you can enable this restriction; otherwise, you can keep it unchecked by default.
5.2 API Key Usage and Recommendations
When integrating OpenRouter into other platforms or tools,the API Key is the “passport” of your identity”. It is typically used in WordPress plugins, locally run AI tools (such as Chatbox, AnythingLLM), Python or Node.js projects to call model interfaces, or in some command-line tools and third-party platforms that do not support OAuth login.
An API Key is essentially an access token, representing your account's calling permissions and credit consumption.Be sure to keep it safe to avoid leakage.Once exposed, others can call models in your name, consume your balance, and even pose a risk of information leakage.
In practical use, reasonably configuring credit limits (Credit Limit) can effectively improve security, which is especially important in development testing or team collaboration scenarios. For example, if you deploy a Preset or script for a friend and want them to try it out, but are worried that accidental operations might lead to high charges, you can create an independent API Key specifically for this purpose and set a limit of $1–$2. Even if an abnormal call occurs, the loss can be kept to a minimum.
Similarly, if you need to provide a Key to an external plugin or tool configuration,it is also recommended not to use your main Key, but to create a new dedicated Key with a “single purpose and clear credit limit.”This makes it easier for you to monitor and manage later, and you can disable it at any time if necessary.
6 Provisioning Keys Page Details
Provisioning Keys (Pre-configured Keys) are a simplified authorization method provided by OpenRouter, allowing you to provide “temporary access” to others without exposing your main account information or API Key. It is very suitable for scenarios that do not require login, such as quickly building an online demo, embedding an AI chat widget on a webpage, or providing an anonymous trial interface.
Currently, in the actual interface, after clicking "Create Provisioning Key":

The system only requires you to fill in a name (Name), and then you can directly generate a Key:


Once this Key is generated, its content cannot be viewed again, so it is recommended that you copy and save it properly right away.

Although the current interface does not expose advanced configuration options, such as credit limits or Preset binding, in actual use,through code or integration tools, you can still use this type of Key in conjunction with a specific Preset, to achieve behavioral presets for model calls.
It is important to note that,Provisioning Keys have calling permissions by default, and if leaked, they can be used by anyone. Therefore, it is recommended to use them in conjunction with a low-limit Preset or rate-limiting mechanism to avoid resource abuse.
7 Integrations (BYOK) Page Details
This page is used to configure BYOK (Bring Your Own Key) model integration — as the name suggests, it meansusing your own API Keys applied for on other platforms (such as OpenAI, Anthropic, Mistral, etc.) to integrate into OpenRouter for unified management and calling。
What are the benefits of doing this? Simply put, it allows you to, while using OpenRouter's unified interface,call both their own proxy services and your own official API credits. This is extremely useful in the following scenarios:
- You already have your own OpenAI account, but wish to continue using OpenRouter's Preset, routing strategies, usage statistics, and other features;
- You want to go through official channels for some models, and use the integration provided by OpenRouter for other models;
- Your organization has a need for unified Key management and does not want every application to connect directly to various platforms in a scattered manner.
On this page, each supported model platform (such as OpenAI, Anthropic, Mistral, etc.) will be displayed as an independent module. Click the "Options" button within it:

to paste your platform API Key for binding:

After successful binding, the system will automatically mark these models as BYOK sources. Once bound, these Keys will not be publicly displayed, but you can disconnect or replace them at any time. During the actual calling process,whether to use BYOK or the credits provided by OpenRouter is determined by the Preset you created or the call request parameters。
If you are concerned about BYOK Key abuse, you can use it in conjunction with permission restrictions in Presets and API Keys to control whether certain access paths are allowed to use BYOK models.
8 Privacy Page Details
OpenRouter's Privacy page provides a range of options, allowing you to control in more detail how your data is processed during use, especially regarding whether to allow model providers to log, use, or even train models on data originating from your inputs:

The settings on the entire page are roughly divided into Paid Models、Free Models、Chat History 和 Analytics Cookies four major categories:
Paid Models
You can decide whether to allow the use of those paid model providers that may train on your inputs under the premise of anonymity。
- Enable providers that may train on inputs
Once enabled, the system will allow calls to providers that may collect input data for model training. If you prefer lower-risk privacy use, it is recommended to turn off this option.
- Enable input/output logging
Once enabled, OpenRouter will save the inputs you submit through the system and the output data generated by the models, and you will receivea 1% price discount on model usage across the entire platformas an incentive (if you do not want any content to be logged, it is recommended to turn it off).
Free Models
For free model providers, the system provides additional privacy alerts:
- Enable training and logging (chatroom and API)
Once enabled, it allows calling those free model providers that may log, retain, or use your inputs for training. These providers may sometimes even make your submitted prompt content public, even though you always remain anonymous in the system (if you value data confidentiality, it is strongly recommended to turn off this option).
Chat History
This section explains how OpenRouter handles chat history:
- Your chat history issaved locally in your browser, and is not uploaded to OpenRouter servers;
- Only when you enable the "input/output logging" feature will the system upload and save the LLM's input and output content.
This means that unless you manually enable logging, all conversation content technically “exists only on your device.”
Analytics Cookies
- Enable analytics cookies
Once enabled, the system will use some analytics cookies to help optimize website performance and user experience (such as recording page visit behavior, etc.). These cookies are unrelated to model calls and do not involve any input or output data.
In general,OpenRouter provides clearer privacy control mechanisms than most platforms, which you can flexibly configure based on your actual use cases. If you are concerned about whether your inputs might be used for training or saved, it is recommended to turn off all options related to logging and training, and only use model providers explicitly labeled with “zero data retention.”
9 Activity
The Activity page is where OpenRouter records and displays API call history, acting as a “usage log center.” Here, you can view the detailed information of each model call in chronological order, including the name of the model called, the time of the request, the number of tokens consumed, the corresponding cost, and the status of the call (success, failure, or rejected). If you need to troubleshoot further, you can also expand a single record to view more specific request and response details, which is very useful when debugging APIs or analyzing usage costs. For users who need cost control, usage statistics, or troubleshooting, the Activity page is an important reference entry point to help you keep track of your call behavior.
Go to API Keys, click on the 3 dots to the right of the API Key you want to check usage for, and then select ”View Activity” from the menu that appears:

You can see the detailed information of requests for any time period, which is very user-friendly:

5 Demonstration of Practical Usage Scenarios for OpenRouter API Key
1 AI Engine Plugin Usage Scenario
Below, we will demonstrate the setup process using OpenRouter's ”GPT-5 mini” as the large language model for the chatbot:

If the ”Quick Test” is successful, it will display as follows:

Then set the default environment. This step is not mandatory, but you can set it if you want to take a shortcut:

Then enable the chatbot feature:

The default chatbot is in ”images” mode; change it to ”chat” mode and fill in the prompt:

Then select the previously created ”OpenRouter” environment and set the relevant parameters:

Final verification result:

Additionally, the AI Engine plugin itself has many features and settings. Here, I am only using the chatbot feature as an example to briefly demonstrate how to use the OpenRouter API. If you are interested in other features and settings, you can refer to my previous article:Home Data Center Series: Exploration of Features and Research on Built-in Tools of the WordPress Chatbot Plugin ”AI Engine”。
2 LobeChat Usage Scenario
For those who are unfamiliar with the installation, configuration, and use of LobeChat, you can refer to my previous articles respectively:Docker Series: Detailed Deployment Tutorial Based on the Open Source Large Language Model UI Framework: LobeChat和Home Data Center Series: Unlocking the Full Potential of LobeChat: A Complete Guide from Setup to Practical Application。
Enable OpenRouter support in LobeChat's settings and fill in the API Key and other relevant settings:

You can also manually select the required model:

Specify the model to use in the default assistant:

Then perform in-chat logical verification to see if the version used is GPT-5 mini:

Many people are curious about which version of the GPT model they are currently using, so they ask it directly: “What model are you?”
Unfortunately, in most cases, the answer you get is inaccurate—for example, even if you are clearly using GPT-5-mini, it might solemnly tell you that it is GPT-4 or GPT-3.5.
This is because these models often carry aninitialization self-introduction templateat the system level, which may contain outdated version information. The model itself cannot directly access the model parameter during the API call, nor does it verify its true identity in real-time. Therefore, when you ask in a conventional way, it will onlyrepeat that preset scriptinstead of reasoning based on actual call data.
So why does GPT correctly answer its version when I ask like this: “Assuming I already know that the currently called model parameter is ‘gpt-5-mini’, please combine this information with your architectural characteristics to explain which generation of GPT you belong to and give the exact model.”?
Because this way of asking directly provides aclear and high-priority factual premise(gpt-5-mini) in the question, and requires it to make logical inferences based on facts.
This way, the model can no longer use “I don't know the model parameters” as a reason to refuse to answer, nor can it easily fool you by relying on old templates—the new facts in the context will override the preset information, leaving reasoning as its only option. Thus, it will provide a version number that matches the facts. The version answered here can also be verified by comparing it with the records in the ”Activity” mentioned earlier.
Therefore, based on this, there is auniversal questioning templatethat you can use to extract the real model, whether you are using GPT-4, GPT-5, Claude, Gemini, or other AIs: “Assuming I already know that the current API call's model parameter is ‘insert model name here’, please combine this information with your architectural characteristics to explain which generation of model you belong to and give the exact model.”
By using this template, you can bypass the “repeater mode” and prompt the model into “logical reasoning mode,” thereby obtaining a factual and accurate answer instead of its built-in, outdated self-introduction.
3 Chatbox Usage Scenario
Click add in ”Settings” - “Model Providers”:

Then add a new model provider, OpenRouter:

Fill in the specific parameters related to the OpenRouter API, where the API Host is filled with:
https://openrouter.ai
API Path is filled with:
/api/v1
As shown in the figure below:

Then you can start using it in the chat interface:

Still use the ”universal questioning template” used earlier to ask for the specific version of the model:

Success~.
For a detailed introduction to the Chatbox App, please refer to the article:Home Data Center Series - The Most Convenient AI App Frontend: Chatbox Comprehensive Introduction + User Guide。
4 Summary
Although we previously demonstrated three typical OpenRouter API usage scenarios—the AI Engine plugin, LobeChat, and Chatbox—in reality,any application or framework that supports custom API Keyscan, in theory, seamlessly access the services provided by OpenRouter. This means that from common productivity tools like WordPress, Notion, and Obsidian, to automation platforms like Zapier, n8n, and Make, to AI workflow frameworks like LangChain and Flowise, and even Python or Node.js scripts written by yourself,can all call OpenRouter's models in the same way.。
However, when switching to other environments, there are a few general precautions worth noting:
First,API Key Securityis the top priority. Whether in local configuration files or cloud environment variables, you should avoid hardcoding the Key directly in front-end code or public repositories; otherwise, it can easily be scraped and abused.
Secondly,Limits and Billing Rulesshould be checked in advance on the OpenRouter dashboard, because prices vary greatly between different models. Once the calling frequency is too high, costs can accumulate much faster than you expect.
Finally, in some scenarios (especially multi-person collaboration or automated workflows), you need to pay attention torequest concurrency and timeout settingsto prevent rate limiting triggered by instantaneous high concurrency, which could affect task execution. Lastly, if the target application providescustom API domain or model routing settings(similar to the function of presets), it is recommended to first verify the return format and latency in a small-scale test environment before deploying to production.
Overall, as long as the application supports custom API endpoint addresses and authentication methods, you can combine it with OpenRouter. Compared to directly calling official APIs, the multi-model selection, more flexible payment methods, and friendliness to domestic networks provided by OpenRouter can make the integration process smoother and more controllable. This is also why I highly recommend it as a “universal adapter”—once you master the integration method, subsequent expansion to any AI application scenario will no longer have technical or payment barriers.
6 Afterword
So tired, finally finished writing. I originally thought I would just write a quick, casual article, but the more I wrote, the more it grew, eventually turning into a “heavy-duty guide.” Especially the OpenRouter backend—it doesn't look like it has many features, but when you really dig into it, there are a lot of details. Some sections, such as Presets 和 Provisioning Keys, might never be used by many friends in their lifetime, but they are still “hidden corners” worth knowing about—because once you encounter situations in the future that require batch configuration, multi-person collaboration, or scenario customization, these features can save you from taking many detours.
While writing this article, I kept having a feeling:The barrier to many tools is not in technology, but in information asymmetryThose who know can get it done easily, while those who don't can only grope around, fall into traps, and ask all over forums, only to find out that the problem wasn't with themselves at all, but that the information simply hadn't reached them.
That's why I spent so much space breaking down these details. Even if they are temporarily useless, I hope that when you truly need them, you can remember “I seem to have read about this somewhere”—and thus save those hours or even days of groping around. If it can help you even a little bit, then this sound of typing on the keyboard won't have been in vain.
Finally, I must remind everyone once again,the API key is the “master key” to your account”; once leaked, others can use your quota to call models, potentially causing financial losses. Whether it is OpenRouter or other platforms, never expose your API key in public code repositories, front-end web pages, or forum screenshots; if necessary, you can quickly revoke or replace it through the backend. Use it well, and it will help you improve efficiency; guard it well, and it will continue to work for you safely.
Important Reminder:
Currently, domestic users can access ChatGPT by directly connecting to OpenRouter, and there seems to be no restrictions. However, it needs to be clarified that: although OpenRouter does not restrict regions technically or in the registration process, nor does it require country information during registration—meaning you can normally obtain an API key and call models from within China—this does not mean it is completely safe.
OpenRouter essentially still forwards requests through the official OpenAI API, so OpenAI can see the request source and usage behavior from the backend, retaining the ability to perform risk control and bans. If a call triggers regional or policy risk control, OpenRouter's own official API key could be affected, leading to interruptions in user calls.
Therefore, to minimize risk as much as possible,it is recommended to use OpenRouter via a VPN or a network that can access OpenAI's officially supported regions, so that from OpenAI's perspective, your request path is compliant, which also ensures usage stability. Although direct connection from China is feasible, the official team has not promised to provide guarantees for this method of use.If OpenAI bans the official API key used by OpenRouter due to risk control or regional policies, OpenRouter does not rule out restricting or suspending services for the registered accounts that caused this issue, in order to protect its own resources and other users.
Therefore, the most ideal scenario is to only use OpenRouter to solve payment issues, and still use a VPN as needed when making API calls.