InferKey

InferKey provides an API service that gives developers access to multiple frontier language models, including Claude and GPT variants, through a single API key. The service is designed to be drop-in compatible with existing OpenAI and Anthropic SDKs by changing the API key and base URL. It operates on prepaid credits with per-token billing at rates stated to be 49-80% below official model pricing.

Developers can start using InferKey by getting a key from their website and changing two lines in their code to point to the new base URL. The process requires pointing existing OpenAI or Anthropic SDKs at InferKey's endpoints. No subscription is needed; users simply purchase prepaid credits to begin.

Pricing is based on prepaid credits with per-token billing at rates stated to be 49-80% below official model pricing. Credits do not expire and can be purchased with card or Alipay. All prices in USD per 1M tokens are billed on actual token usage, and your rate is fixed at purchase time.

The service offers 16+ frontier models behind one key, including Claude Fable 5, Opus 4.8, Sonnet 5, GPT-5.5, and GPT-5.6. Official models are served unmodified, and savings are measured and published as up to 80% below official pricing. The system supports vision and PDF input across every model.

Users can send images and PDFs to the models, as the service explicitly supports vision and PDF input for every model. Image generation for gpt-image-2 is billed by the generated image's size tier rather than quality. Size tiers include 1K (≤1024²) at $0.0177, 2K (≤2048²) at $0.0264, and 4K (>2K) at $0.0354.

The platform supports drop-in compatibility with existing OpenAI and Anthropic SDKs, allowing developers to switch with minimal code changes. Users can access the service via two base URLs depending on the model family they are using. The API key and base URL are the only changes required to integrate the service.

The website states that all prices in USD per 1M tokens apply to new purchases, and the rate is fixed at the time of purchase. This means historical rates do not change for existing credits, and future purchases are locked in at the current price. The pricing table includes specific input, output, and cache costs for each model variant.

The service claims to offer up to 80% savings compared to official pricing, with specific examples like gpt-5.6-sol showing an 81% reduction. Claude Fable 5 offers up to 49% savings, while gpt-5.4-mini can reach up to 80% savings. These figures are based on official model responses served unmodified.

The platform provides transparent per-token pricing with a clear breakdown of input, output, and cache costs for each model. For example, claude-sonnet-5 has an input cost of $1.53 per 1M tokens and an output cost of $7.68. The pricing table is available on the website for verification.

The service is designed for developers who ship, offering a drop-in compatible, radically cheaper, and honest solution for accessing frontier models. It targets users who need access to multiple models without managing multiple API keys or subscriptions. The platform aims to simplify the integration process for developers.

The website mentions that the service is built for developers who ship, emphasizing speed and ease of integration. It provides a single API key for multiple models, reducing the complexity of managing different providers. The platform focuses on making it easy to switch between models without code changes.

The service operates without a subscription model, relying instead on prepaid credits that never expire. This allows users to pay only for what they use, with no ongoing fees. The system supports both card and Alipay payments for global accessibility.

The platform provides a single API key for access to 16+ frontier models, including both Claude and GPT variants. This consolidation simplifies the development process by allowing users to manage multiple models through one interface. The service is designed to be a one-stop solution for developers needing various model capabilities.

You can reach InferKey by e-mail at support@inferkey.dev.

The official website of InferKey is inferkey.dev.