FAQ & Solutions

Oken is a price-comparison and monitoring platform focused on AI-model token pricing and performance. We aggregate API pricing and route-latency data from every major official channel, cloud provider, and third-party reseller — helping developers quickly find the AI-model API with the best value, the fastest speed, and the best fit for their use case.

• Network-wide price comparison:Cuts through the information gap so you can compare the same model's price across channels at a glance, with an “official benchmark price” as your reference.
• Performance monitoring:Real-time probing of key metrics like each route's connectivity and time-to-first-byte.
• One-stop filtering:Filter by price, provider type, context length, and other fine-grained criteria to zero in on the best option fast.

An AI model's running cost is driven mainly by compute. Typically, reading the prompt you send it (input) is relatively cheap, while generating new text (output) takes more compute and therefore costs more. Oken breaks these dimensions out so you can estimate costs precisely.

• Publisher (Official):The official API provided directly by the model maker — e.g., OpenAI or DeepSeek.
• Cloud Provider:Officially authorized cloud platforms that host model services — e.g., Alibaba Cloud, Volcano Engine, AWS. (Models like Qwen and Doubao are also offered through cloud providers.)
• Aggregator / Reseller:API distribution platforms built by third-party developers. They usually bundle many models together, have a low top-up minimum, and are flexible to use — but route stability and billing multipliers vary from site to site.

To make management easier, some third-party resellers bundle routes for different models together, or use inconsistent internal naming conventions. Wherever possible, Oken cleans up and re-tags these routes so that what you see reflects the real routes for the current model.

The prices shown on Oken are usually the channel's base list price. Some overseas channels or China-based enterprise-grade cloud providers may add extra tax at settlement, depending on your registration location or invoicing requirements. For specifics, refer to the final provider's checkout page after you're redirected there.

Many third-party top-up platforms use a fixed internal exchange rate (e.g., 1 USD = 7 RMB, or 1 USD = 10 RMB) to hedge against currency swings or simply to keep their billing logic simple. The prices Oken displays have already been converted into intuitive, comparable units wherever possible.

Some advanced models (e.g., Claude 3.5, DeepSeek-V3) support Prompt Caching. When you repeatedly send large amounts of identical background content (like very long documents or system prompts), the system caches it. Reading from cache uses far less compute than recomputing from scratch, so those tokens usually get a steep discount.

• Context Window:The total amount of information a model can “hold in mind” during a single conversation (including your questions, the conversation history, and the model's replies).
• Max Output Tokens:The maximum length of new content a model can generate in a single reply. (For example: a model might read a 1-million-token financial report, but in one go it can only write a summary of up to 128k tokens.)

Price data is updated regularly through a mix of automated scraping and manual review. As for latency:Oken's probes are currently deployed at fixed nodes, so the figures you see are objective measurements taken from our servers. Because your actual client sits in a different geographic location, real-world network overhead may differ. We're planning multi-node testing (across regions such as China, overseas, and others) to deliver more useful reference data.

Once you've found a route you like in the list, click the channel's name or its action button and you'll be taken directly to that provider's official website. You'll need to register and top up on that platform.

No matter which channel you use, you usually just need to change two key settings in your code or third-party client (like Chatbox or NextChat):

• Base URL:Replace it with the address provided by the platform you bought from.
• API Key:Replace it with the dedicated key you generated on that platform. Once you've swapped both, you can use it seamlessly — just like the official API.