claude-opus-5 API — per-token pricing and code examples

Checking statusPrice version 4 · in effect since Oct 31, 2026

Price per 1M tokens

InputOfficial price $5.00$1.15
Cache readOfficial price $0.50$0.115
Cache writeOfficial price $6.25$1.15
Cache write 1 hOfficial price $10.00$1.15
OutputOfficial price $25.00$5.75

Minimum per request$0.00002

Cache rates apply when the provider reports cache tokens.

77% off

About this model

claude-opus-5 is served through OpenAI Chat Completions, OpenAI Responses, Anthropic Messages. You pay $1.15 per 1M input tokens and $5.75 per 1M output tokens. That is 77% below the official price of $5.00 and $25.00. Cache reads cost $0.115 per 1M tokens when the provider reports cached tokens. It is included in budget packs from $99.00 up.

Where Claude Opus 5 fits

Claude Opus 5 is the Opus model Anthropic released on July 24, 2026. It is now listed as legacy, with a retirement date of not sooner than July 24, 2027, and Anthropic points new projects to Claude Opus 5.5. Its knowledge runs to May 2026. Opus 5 accepts text and images, writes text, and offers a 1M-token context window with up to 128K output tokens in one synchronous call.

Opus 5 is a good fit when you need the one Claude 5 Opus that still lets you switch thinking off, or when a workflow forces a specific tool on every call. Both of those stopped working on Opus 5.5, so Opus 5 is the natural place to stay while you rework that code.

Thinking on by default, with an off switch

On Opus 5, a request without a thinking field already runs adaptive thinking. Unlike the later Opus 5.5, you can still turn it off: Anthropic says Opus 5 accepts thinking: {type: "disabled"} at effort high or below. At xhigh or max, thinking cannot be turned off, and the disabled setting is rejected with 400. The enabled mode with budget_tokens and the between_tools mode are both rejected on this model.

Effort defaults to high and accepts all five levels, low through max. Anthropic's advice is to start at high and move up to xhigh for demanding coding and agentic work. Per-message effort is available as a beta, except on Bedrock. One honest caveat from Anthropic's effort guide: on Opus 5, changing effort does not reliably shorten responses. If you need shorter answers, ask for brevity in the prompt or cap max_tokens rather than counting on low alone.

Thinking blocks come back with display set to omitted by default, so the block text is empty and only a signature is returned. Forced tool use is still allowed on Opus 5, and structured outputs are supported as well.

What it costs on Cent AI

Opus 5 is billed per token against your prepaid USD balance, at the four rates in the price box; cache writes are billed at the input rate. Because thinking runs unless you disable it, most Opus 5 requests include reasoning tokens, and those are charged as output even though the default display hides their text. For high-volume, simple calls, thinking: {type: "disabled"} at low or medium effort is the most direct way to keep output small.

Caching starts at 512 tokens on Opus 5. The default lifetime is five minutes, and you can request a one-hour lifetime with ttl: "1h" in cache_control. Cent AI bills cache reads and writes only from the cached token counts in the model's usage report. Switching effort or thinking mode mid-conversation throws away cached message blocks, so a fixed setting keeps the cache read rate working for you.

Using it from your tools

You can call Opus 5 from /v1/messages, /v1/chat/completions or /v1/responses. Cent AI does not change your parameters silently. Expect a 400 for any non-default temperature, top_p or top_k, and from API version 2026-10-21 even a default value is refused, so remove those fields. A prefilled last assistant turn is also refused.

This /v1/messages request turns thinking off for a quick lookup:

{
  "model": "claude-opus-5",
  "max_tokens": 1024,
  "thinking": { "type": "disabled" },
  "output_config": { "effort": "low" },
  "messages": [{ "role": "user", "content": "Which HTTP status means rate limited?" }]
}

Specification

Context window
—
Input and output
Text
Endpoints
/v1/chat/completions, /v1/responses, /v1/messages
Smallest pack
$99.00

Latency and error rate per model are on the status page. status.cent-ai.dev

What a request costs

Open the full calculator

Cost per request

90% is an assumption you can change, not a promise. Cache is billed only from cache tokens the model provider reports.

See how many such requests it covers.

Estimate

claude-opus-5
Fresh input99,000 tok
Cache read891,000 tok
Output10,000 tok

Per request$0.2738
At official pricesAt official prices $1.1905

Total$0.2738

An estimate at price version 4. Your bill comes from the usage the provider reports for each request.

Save $0.9166

Code

curl https://api.cent-ai.dev/v1/chat/completions \
  -H "Authorization: Bearer $CENT_AI_KEY" \
  -H "Content-Type: application/json" \
  -d '{"model": "claude-opus-5", "messages": [{"role": "user", "content": "Hello"}]}'

Set CENT_AI_KEY to a key from the dashboard. OpenAI SDKs use https://api.cent-ai.dev/v1; Anthropic SDKs use https://api.cent-ai.dev.

Price history

Version 4, in effect since Oct 31, 2026.

  1. Effective

    per 1M tokens

    Inputold $1.15 new $1.15
    Cache readold $0.115 new $0.115
    Cache writeold $1.4375 new $1.15
    Cache write 1hold $2.30 new $1.15
    Outputold $5.75 new $5.75
  2. Effective

    per 1M tokens

    Inputold $0.85 new $1.15
    Cache readold $0.085 new $0.115
    Cache writeold $1.0625 new $1.4375
    Cache write 1hold $1.70 new $2.30
    Outputold $4.25 new $5.75
  3. Effective

    per 1M tokens

    Inputold $0.95 new $0.85
    Cache readold $0.095 new $0.085
    Cache writeold $1.1875 new $1.0625
    Cache write 1hold $1.90 new $1.70
    Outputold $4.75 new $4.25

Compare with

All models against official prices

Questions

How much does claude-opus-5 cost on Cent AI?

$1.15 per 1M input tokens and $5.75 per 1M output tokens. Cache reads cost $0.115 and cache writes $1.15 per 1M tokens. Each request is rounded up once.

Is there a minimum charge?

A successful request costs at least $0.00002. A request that ends early, because the client left or the output was cut off, pays only for its tokens.

How do I pay?

Top up a prepaid USD balance with QRIS or USDT/USDC, or buy a budget pack once. Nothing renews automatically.

What if the provider does not report usage?

The request is billed from an estimate and marked estimated on the Requests page. An estimate never contains cache tokens.

Which endpoint do I call?

/v1/chat/completions, /v1/responses, /v1/messages. The OpenAI base URL is https://api.cent-ai.dev/v1.

Try claude-opus-5 with one key

One prepaid balance for every model. Pay with QRIS or USDT/USDC.

Start free