# Claude Opus 5 (Fast)

Claude Opus 5 with fast mode switched on. Claude Opus 5 (Fast) is the same model given more compute per request, so output arrives sooner without dropping to something smaller.

- **Model ID:** `anthropic/claude-opus-5-fast`
- **Type:** chat
- **Providers:** anthropic
- **Context window:** 1,000,000
- **Maximum output tokens:** 128,000
- **Pricing:** $10/1M input tokens, $50/1M output tokens
- **Canonical page:** https://vercel.com/ai-gateway/models/claude-opus-5-fast

## Supported parameters

Detailed capability metadata has not been reported for this model.

## Example

```ts
import { streamText } from 'ai'

const result = streamText({
  model: 'anthropic/claude-opus-5-fast',
  prompt: 'Why is the sky blue?'
})
```

## About

Claude Opus 5 (Fast) is Claude Opus 5 served in Anthropic's fast mode. Fast mode is not a different model and not a smaller one: it is a configuration of the same model that allocates more compute per request, so responses arrive sooner at identical quality and with identical capabilities.

On Anthropic's own API you opt in per request with a speed setting and a beta header, and the response reports which tier served it. Through AI Gateway you select it by model id instead, calling `anthropic/claude-opus-5-fast` directly.

Everything that defines Claude Opus 5 carries over, including the 1M tokens context window, reasoning on by default with configurable effort, tool calling, structured output, and streaming. AI Gateway publishes live latency and throughput metrics on this page, so you can compare against the standard model rather than working from a quoted multiplier.

Interactive work is where this earns its price. Rapid iteration, live debugging, and agent loops that chain many calls all shorten at every step, and the saving accumulates across a run. A single background completion rarely justifies it.

You can integrate Claude Opus 5 (Fast) through AI SDK, Chat Completions API, Responses API, Messages API, or other API formats, from TypeScript or Python.

## What to consider

Fast mode is a research preview with gated access, so confirm your account can reach it before you design around it. It also carries a rate limit separate from the standard model's. Exceeding that limit returns a 429 with a retry-after header, and because the limit replenishes continuously the wait is usually short, but your retry logic should still handle it.

Provider coverage is narrower than Claude Opus 5, and not for the usual reason. Fast mode is served through Anthropic's own API rather than the cloud resellers, so the providers listed on this page will be a shorter set than the standard model's.

Pricing is a premium over Claude Opus 5 and the multiplier applies across the whole context window, including requests past 200K input tokens. It also stacks with prompt caching and data residency multipliers, so the effective rate compounds rather than replacing them. Check the pricing panel on this page before moving a workload across.

Quality is not a reason to choose Claude Opus 5 (Fast). The model is the same, so a prompt that answers poorly on Claude Opus 5 answers the same way here, only sooner. Change models rather than speed when the output is the problem.

## When to use

### Best For

- **Rapid iteration** where a developer is waiting on each response and latency is felt directly
- **Live debugging** on interactive surfaces that stall while a standard response streams
- **Long tool-call chains** in which a saving on every call accumulates across the run
- **Unchanged output quality** when you want Claude Opus 5 behaviour delivered sooner

### Consider Alternatives When

- **Batch and background work** carries the load, where Claude Opus 5 costs less and nobody is waiting
- **Cost per token dominates** the decision, since the premium applies across the full context window
- **Better answers** are the real requirement, where the same model will not help
- **Cloud provider routing** is required, since fast mode is served through Anthropic's own API

## Best for

- **Rapid Iteration:** Responses a developer waits on directly
- **Live Debugging:** Interactive surfaces that stall on a standard response
- **Long Tool-Call Chains:** Runs where a saving on every call accumulates
- **Unchanged Output Quality:** Claude Opus 5 behaviour delivered sooner

## Consider alternatives

- **Batch And Background Work:** Claude Opus 5 costs less when nobody is waiting
- **Cost-Driven Workloads:** The premium applies across the full context window
- **Quality Problems:** The same model returns the same answer, only sooner
- **Cloud Provider Routing:** Fast mode is served through Anthropic's own API

## Frequently asked questions

### What is fast mode?

A configuration of Claude Opus 5 that allocates more compute per request so output arrives sooner. It is not a different model and not a smaller one, and quality and capabilities are identical.

### Will Claude Opus 5 (Fast) give better answers than Claude Opus 5?

No. It is the same model, so the same prompt produces the same quality of answer. Choose Claude Opus 5 (Fast) for latency, and a different model if quality is the problem.

### How do I use fast mode?

On AI Gateway, call `anthropic/claude-opus-5-fast` as its own model id. On Anthropic's API directly you opt in per request with a speed setting and a beta header, and the response reports which tier served it.

### Why do fewer providers serve Claude Opus 5 (Fast)?

Fast mode runs on Anthropic's own API rather than the cloud resellers, so the provider list is shorter than Claude Opus 5's. See the providers shown on this page.

### How much more does fast mode cost?

It is a premium over Claude Opus 5, and the multiplier applies across the whole context window, including requests past 200K input tokens. It stacks with prompt caching and data residency multipliers rather than replacing them. See the pricing panel on this page.

### Is there a separate rate limit for Claude Opus 5 (Fast)?

Yes, separate from the standard model's. Exceeding it returns a 429 with a retry-after header. The limit replenishes continuously, so the wait is usually short, but handle the retry in your client.

### Is fast mode generally available?

It is a research preview with gated access. Confirm your account can reach it before designing a workload around it.

### What is the context window for Claude Opus 5 (Fast)?

The context window is 1M tokens, with up to 128K tokens per response, matching the standard model.

### Does Claude Opus 5 (Fast) support Zero Data Retention?

Yes, Zero Data Retention is available for this model. Zero Data Retention is offered on a per-provider basis. See https://vercel.com/docs/ai-gateway/capabilities/zdr for details.

## Links

- [Model page](https://vercel.com/ai-gateway/models/claude-opus-5-fast)
- [AI Gateway documentation](https://vercel.com/docs/ai-gateway)
- [Provider model documentation](https://docs.anthropic.com/claude/docs/models-overview)