Anthropic: Claude Opus 5.5

Claude model for long-running agentic coding and knowledge work

claude-opus-5-5

AvailableAnthropicChatCached input 95% less

Input / Output-70%: $4.00 / $20.00$1.20 / $6.00
Context: 1M
Released: Sep 23, 2026
Max output: 128K
Modalities: VisionChat
Capabilities: Tool usePrompt CacheReasoning

About Claude Opus 5.5

Claude Opus 5.5 is Anthropic's current Opus model, released on September 22, 2026, for long-running agentic coding and knowledge work. Anthropic says it performs at Fable 5.1 level on most tasks, produces output much faster than Opus 5, and uses fewer tokens and turns for equal work. Adaptive thinking is always on and cannot be disabled.

Where it works well

Anthropic reports it ahead of both Opus 5 and Fable 5.1 on Terminal-Bench 4.0, its agentic terminal-coding benchmark.
Anthropic reports strong results on OSWorld 2.1, so desktop and browser automation is a good fit.
Stays on task over very long runs; one customer described an agent working more than 18 hours on a service architecture.
Medium is the default effort, and Anthropic says medium often matches earlier models at their maximum setting.

When to choose another model

Thinking cannot be turned off, and forced tool use returns an error, which breaks code written for Opus 5.
Text between tool calls arrives in thinking blocks that are empty by default, so apps that stream progress text go quiet until you change the display setting.
For the hardest reasoning, Anthropic still points to Fable 5.1, and Sonnet 5.5 answers faster on simpler jobs.

Getting started

Create API key
Create a key in Console, then use it with every model on the platform.

Send your first request

Copy the example for your language and run it against the endpoint.

Text to textMessages API

POST/v1/messages

API format:

curl https://api.tokenlab.sh/v1/messages \
  -H "Content-Type: application/json" \
  -H "x-api-key: sk-xxx" \
  -H "anthropic-version: 2023-06-01" \
  -d '{
    "model": "claude-opus-5-5",
    "max_tokens": 1024,
    "messages": [
      {"role": "user", "content": "Hello!"}
    ]
  }'

Pricing

The Verified price applies to Verified, which costs less on most models. The Official price is the model maker's published price and applies to the more reliable Official route. Auto bills the route that completes the request.

Default spec

per 1M tokens

Official price: Input $4.00 / Output $20.00 / Cache read $0.20 / Cache write $5.00
Verified price: Input $1.20 / Output $6.00 / Cache read $0.06 / Cache write $1.50
Discount: -70%

Prompt cache pricing

Cache read

Official price: $0.20
Verified price: $0.06
Discount: -70%

Cache write

Official price: $5.00
Verified price: $1.50
Discount: -70%

Tool fees

Charged per call, only when the model uses the tool.

Web search

Official price: $0.01/search
Verified price: $0.003/search
Discount: -70%

	Official priceper 1M tokens	Verified priceper 1M tokens	Discount
Default spec	Input $4.00 / Output $20.00 / Cache read $0.20 / Cache write $5.00	Input $1.20 / Output $6.00 / Cache read $0.06 / Cache write $1.50	-70%
Prompt cache pricing
Cache read	$0.20	$0.06	-70%
Cache write	$5.00	$1.50	-70%
Tool feesCharged per call, only when the model uses the tool.
Web search	$0.01/search	$0.003/search	-70%

Usage & activity

Success rate is the share of requests that completed. Latency is how long a full response takes; P95 means 95% of requests finished within that time.

Usage & availability

Last 24 hours

Requests
Success rate
P95 latency
Total tokens

Model performance

Last updated Oct 16, 2026, 11:11 AM UTC

30-day success rate: 96.4%
7-day median latency: 6.0sn=19,882
7-day P95 latency: 27.5sn=19,882
30-day requests: 20K+
Last active: 12 seconds ago

Data is based on aggregate user requests, excluding status checks.

Open in Console

Open Claude Opus 5.5 in Console with a prompt ready to edit or send.

Help me try claude-opus-5-5 with a short message at /v1/messages. Show the reply, latency, and cost.

Use cases

Best for

Reasoning
Vision

Large codebase migrations
Run a framework or language migration over a big repository, with the agent planning batches, converting files, and fixing the failing tests.
Computer-use automation
Operate desktop and browser software through screenshots for back-office tasks that have no API.
Multi-hour knowledge work
Assign a long research or operations task, such as auditing an architecture or compiling a report from many documents, and check back on the result.

Prompt examples

Migrate this monorepo from Jest to Vitest, keep CI green after each package, and summarize anything you changed beyond the config.

Open the vendor portal, download last month's invoices, and enter each total into the reconciliation sheet.

Audit our service dependency graph, find single points of failure, and propose a prioritized hardening plan.

This model has conditional pricing. Monthly token totals alone cannot produce a reliable estimate; use the detailed pricing for the request specification, cache, and applicable time window.

FAQ

What is new in Claude Opus 5.5 versus Opus 5?

Anthropic reports much faster output, lower cost on typical workloads, higher scores such as on Terminal-Bench 4.0, and stronger prompt-injection resistance. It also changes API behavior: thinking is always on and forced tool use is rejected.

Can I disable thinking on Opus 5.5?

No. Adaptive thinking is always on, and you control depth with the effort parameter, which defaults to medium and also offers low, high, and xhigh. Forced tool use also returns an error, so remove tool_choice settings that force a specific tool.

Is Claude Opus 5.5 better than Fable 5.1?

Anthropic says Opus 5.5 performs at Fable 5.1 level on most tasks and beats it on several reported benchmarks, such as Terminal-Bench 4.0. Fable 5.1 remains the choice when your evaluations at higher effort still fall short.

Does Opus 5.5 work with images?

Yes. It accepts text and images and answers in text, which is what makes its computer-use scores possible: it reads the screen, acts, and checks the result. It does not generate images.

How much does Claude Opus 5.5 cost?

On TokenLab, Claude Opus 5.5 costs Input $1.20 / Output $6.00 per 1M tokens. The pricing table above shows the full breakdown. Rates depend on the billing unit, specification, and usage. Compare matching conditions in the model's detailed pricing; a single rate does not determine the total cost.

What are the context window and output limit of Claude Opus 5.5?

Claude Opus 5.5 accepts up to 1,000,000 tokens of context and returns up to 128,000 tokens in one response.

Which endpoint should Claude Opus 5.5 use?

Use https://api.tokenlab.sh/v1/messages for Claude Opus 5.5. The request example below shows the matching code shape.

Which operations does Claude Opus 5.5 support?

Claude Opus 5.5 supports Text to text. Select an operation above to see its endpoint and request example.