Claude Opus 5.5 is out: 1M context, always-on thinking, $4/$20

What you get

A quick upgrade path to Claude Opus 5.5, released September 22, 2026: 1M-token context, 128K max output, always-on adaptive thinking, priced at $4 / $20 per million input / output tokens.

Prerequisites

A Claude API key in ANTHROPIC_API_KEY
The anthropic Python SDK (or @anthropic-ai/sdk for TypeScript)

Steps

Update the SDK.

Shell

pip install -U anthropic

Switch the model ID to claude-opus-5-5 (save as quickstart.py).

Python

import anthropic

client = anthropic.Anthropic()

message = client.messages.create(
    model="claude-opus-5-5",
    max_tokens=1000,
    messages=[{"role": "user", "content": "Summarize what changed in my last deploy."}],
)

for block in message.content:
    if block.type == "text":
        print(block.text)

Check the breaking changes if you are coming from Claude Opus 5:

Thinking can't be disabled; steer depth with the effort parameter (default medium).
Forced tool use returns an error.
Thinking blocks are tied to the model and conversation that produced them.
The older computer_20251124 computer use tool is not accepted on the Claude API and Google Cloud.
Text between tool calls now comes back in thinking blocks, empty at the default display setting.

Test it

Shell

python quickstart.py

The response model field should read claude-opus-5-5. Watch cost: cache reads are $0.20 / MTok and Batch API requests get 50% off.