How CheapAI Works
A transparent explanation of the model, methodology, what CheapAI actually does, and where the limitations are. If something here is unclear, contact us directly.
The Business Model
Volume aggregation
CheapAI pools API usage across many developers. Higher aggregate volume unlocks better rates with providers than an individual developer would receive on a pay-as-you-go plan. These savings are passed to customers.
Shared proxy infrastructure
Requests are routed through CheapAI's OpenAI-compatible proxy endpoint. The proxy translates your call to the upstream provider API. All providers currently supported: Anthropic, Google, DeepSeek, and OpenAI-compatible models.
Lean operations
CheapAI is a small developer team with lean infrastructure costs. Unlike enterprise API portals, there is no large sales team or enterprise overhead. This keeps operational costs low and allows deeper discounts.
How Your Request Travels
- Your app sends a standard OpenAI-compatible API request to
https://cheapai-netifly-app.up.railway.app/v1/chat/completions - CheapAI's proxy receives the request, validates your API key, and routes it to the appropriate upstream provider based on the requested
modelparameter. - The upstream provider (Anthropic, Google, DeepSeek, or OpenAI-tier) processes the request and returns a response.
- CheapAI's proxy forwards the response to your application, in the same format as the official OpenAI API.
- Tokens consumed are counted against your purchased balance or plan. You are billed at CheapAI's discounted rates, not official provider rates.
The round-trip adds a small proxy overhead (typically <50ms tested, though not guaranteed). The response format is identical to the official API for all supported parameters.
Data Handling
- CheapAI does not log or store the content of your prompts or completions beyond what is required for per-request billing validation.
- Request metadata (timestamp, model, token count, key ID) is retained for billing purposes.
- Upstream providers (Anthropic, Google, DeepSeek, OpenAI) receive your prompts as part of normal API operation and their own data policies apply. CheapAI does not provide any additional data to providers beyond what a direct API call would.
- Do not send personally identifiable information, secrets, or regulated data through a shared proxy — this applies to any AI API proxy service, including CheapAI.
See our privacy policy and terms of service for full legal details.
Scope & Limitations
- Not a direct provider relationship: CheapAI is a proxy, not a direct customer of Anthropic, Google, or OpenAI on your behalf. Your prompts travel through CheapAI's infrastructure before reaching the provider.
- Upstream availability: If a provider API is down or degraded, CheapAI's routing for that provider will also be affected. CheapAI is not able to guarantee provider uptime.
- GPT-5.x model IDs: CheapAI's GPT-5.x model names (GPT-5.4, GPT-5.3 Codex, etc.) are internal routing labels. No directly corresponding public OpenAI model IDs were found on openai.com/api/pricing as of March 2026. See our pricing tracker for current verification status per model.
- Gemini preview models: Gemini 3.1 Pro Preview and Flash-Lite Preview are preview-tier models. Pricing and availability may change before stable release per Google's preview terms.
- Rate limits: CheapAI's shared infrastructure applies fair-use rate limiting. Very high-throughput use cases should contact support before purchase to confirm suitability.
- Feature parity: Most OpenAI SDK parameters work identically. Some provider-specific advanced features (e.g. extended_output for Claude, Gemini grounding) may behave differently via proxy. See compatibility matrix.