Documentation Index
Fetch the complete documentation index at: https://docs.gpuhub.com/llms.txt
Use this file to discover all available pages before exploring further.
You can call supported models through a compatible API, monitor usage, and pay based on actual consumption.
Model availability, context length, and final prices may vary by model. Use the prices displayed in the console as the source of truth before starting production traffic.
Supported Model Types
AI Gateway supports multiple categories of models for common AI workloads.
| Model Type | Typical Use Cases | Billing Unit |
|---|
| Language models | Chat, reasoning, coding, summarization, tool calling, and structured output | Input and output tokens |
| Long-context language models | Document analysis, large codebase analysis, and retrieval-augmented generation | Input and output tokens, sometimes tiered by context size |
| Claude-series models | Chat, coding, reasoning, and workflows that use prompt caching | Input tokens, output tokens, cache writes, and cache reads |
Pricing
For each model, use the unit price displayed in the console as the source of truth.
Language Models
General language models are billed by token usage.
| Usage Type | Unit Price |
|---|
| Input tokens | $XX.XX / 1M tokens |
| Output tokens | $XX.XX / 1M tokens |
Total cost =
input tokens / 1,000,000 * input token price
+ output tokens / 1,000,000 * output token price
Tiered Context Pricing
Some language models use tiered pricing based on the number of input tokens in a request. For example, requests within a shorter context window may use one price, while requests with a larger context window may use a higher price.
| Context Tier | Input Token Price | Output Token Price |
|---|
| Standard context | $XX.XX / 1M tokens | $XX.XX / 1M tokens |
| Extended context | $XX.XX / 1M tokens | $XX.XX / 1M tokens |
Total cost =
input tokens / 1,000,000 * tiered input token price
+ output tokens / 1,000,000 * tiered output token price
Claude-Series Prompt Caching
Claude-series models may include prompt caching charges in addition to normal input and output token charges.
| Usage Type | Unit Price |
|---|
| Input tokens | $XX.XX / 1M tokens |
| Output tokens | $XX.XX / 1M tokens |
| Cache write, 5-minute TTL | $XX.XX / 1M tokens |
| Cache write, 1-hour TTL | $XX.XX / 1M tokens |
| Cache read | $XX.XX / 1M tokens |
Total cost =
input tokens / 1,000,000 * input token price
+ output tokens / 1,000,000 * output token price
+ 5-minute cache write tokens / 1,000,000 * 5-minute cache write price
+ 1-hour cache write tokens / 1,000,000 * 1-hour cache write price
+ cache read tokens / 1,000,000 * cache read price
Usage Notes
- Check the console for the latest model list, supported parameters, context limits, and real-time prices.
- For language models, estimate cost from both input and output tokens. Long prompts, retrieved documents, and tool results can increase input token usage.
- For cached prompts, cache writes and cache reads may appear as separate billable usage types.
- Use rate limits and budget controls when integrating AI Gateway into production applications.
