Claude Opus 5

Anthropic’s flagship: near-Fable-5 intelligence at Opus 4.8 pricing, with a 1M context window and thinking on by default.

Specifications

Item	Value
Model ID	`claude-opus-5` · `claude-opus-5-thinking`
Vendor	Anthropic
Vendor release	2026-07-24
Available on APIYI	2026-07-25
Knowledge cutoff	Not disclosed
Input modalities	Text, Image
Output modalities	Text
Context window	1,000,000 tokens
Max output	128,000 tokens
Billing	Usage-based

Pricing

Prices in USD per 1M tokens ($/1M).

Input	Cached input	Output
$5	$0.5	$25

The table shows list prices. Recharge promotions and group discounts stack; the console reflects the actual charge in real time. See Pricing and Recharge promotions.

Endpoints

Endpoint	Path	Supported
`OpenAI Chat Completions`	`POST /v1/chat/completions`	✅
`OpenAI Responses`	`POST /v1/responses`	—
`Anthropic Messages`	`POST /v1/messages`	✅
`Gemini Generate Content`	`POST /v1beta/models/{model}:generateContent`	—
`Image Generations`	`POST /v1/images/generations`	—
`Embeddings`	`POST /v1/embeddings`	—

Billing groups

Group	Multiplier	Notes
`ClaudeCode`	0.95×	95% of list price
`Default`	1×	List price
`SVIP`	1×	List price

Some groups carry extra discounts, stackable with recharge bonuses. See Tokens and groups.

Supported features

Feature	Supported
Streaming	✅
Tool calling	✅
Structured outputs	✅
Vision	✅
Prompt caching	✅
Extended thinking	On by default
Web search	✅

Variants

Model ID	Notes
`claude-opus-5`	Standard call. Omitting the thinking parameter runs adaptive thinking.
`claude-opus-5-thinking`	Forced-thinking variant; pricing and endpoints identical to the standard model.

Example request

The example below calls claude-opus-5 through OpenAI Chat Completions (/v1/chat/completions). Point base_url at https://api.apiyi.com/v1 — everything else matches the official API.

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

Read the key from an environment variable, never hard-coded. In production, issue separate tokens per use case so you can revoke and attribute usage individually.

Launch announcement

Background, benchmarks and migration notes for Claude Opus 5

Claude API Basics

Parameters, usage and best practices

Effort & Thinking

Parameters, usage and best practices

Prompt Caching

Parameters, usage and best practices

Claude Sonnet 5

Details for a model in the same family

Claude Fable 5

Details for a model in the same family

Claude Opus 4.8

Details for a model in the same family

Model pricing directory

Live pricing, endpoints and groups for all 308 models

Specs on this page are maintained by hand in models/data/model-details.json; pricing and endpoints come from the live pricing API, updated 2026-07-31 01:30 (UTC+8), specs last verified 2026-07-31.

Pricing Table

OpenAI

Anthropic

Google

xAI

Alibaba

ByteDance

Moonshot

Zhipu

MiniMax

Specifications

Pricing

Endpoints

Billing groups

Supported features

Variants

Example request

Launch announcement

Claude API Basics

Effort & Thinking

Prompt Caching

Claude Sonnet 5

Claude Fable 5

Claude Opus 4.8

Model pricing directory

​Specifications

​Pricing

​Endpoints

​Billing groups

​Supported features

​Variants

​Example request

​Related docs

Launch announcement

Claude API Basics

Effort & Thinking

Prompt Caching

Claude Sonnet 5

Claude Fable 5

Claude Opus 4.8

Model pricing directory

Specifications

Pricing

Endpoints

Billing groups

Supported features

Variants

Example request

Related docs