Anthropic: Claude Opus 4.6

High-end Claude for difficult coding, planning, and slower expert reasoning

claude-opus-4-6

AvailableAnthropicChatCached input 90% less

Input / Output-70%: $5.00 / $25.00$1.50 / $7.50
Context: 1M
Max output: 128K
Modalities: VisionChat
Capabilities: Tool usePrompt CacheReasoning

About Claude Opus 4.6

Claude Opus 4.6 is Anthropic's Opus-tier model from February 2026, built for difficult coding, planning, and long agent tasks. It introduced adaptive thinking with four effort levels, context compaction for long-running sessions, and a long context window in the Claude 4 line. It is now a legacy model on Anthropic's lineup, kept for workloads already tuned to it.

Where it works well

Breaks ambitious coding projects into concrete steps and carries them through, which early testers cited in Anthropic's launch post.
Adaptive thinking lets the model decide when deeper reasoning is worth it, with low, medium, high, and max effort levels to steer cost.
Context compaction summarizes older turns automatically, so long agent sessions do not stall when the history grows.
Anthropic reports leading results on Terminal-Bench 2.0 for agentic coding and on Humanity's Last Exam at launch.

When to choose another model

Its reliable knowledge cutoff is older than the Opus 4.7 and later models, so recent frameworks are less familiar.
Anthropic now marks it legacy and recommends Opus 5.5, which performs better on current workloads.
The manual budget form of extended thinking is deprecated on this model; plan around adaptive thinking and effort instead.

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-4-6",
    "max_tokens": 1024,
    "messages": [
      {"role": "user", "content": "Hello!"}
    ]
  }'

Pricing

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

Default spec

per 1M tokens

Official price: Input $5.00 / Output $25.00 / Cache read $0.50 / Cache write $6.25
TokenLab price: Input $1.50 / Output $7.50 / Cache read $0.15 / Cache write $1.88
Discount: -70%

Prompt cache pricing

Cache read

Official price: $0.50
TokenLab price: $0.15
Discount: -70%

Cache write

Official price: $6.25
TokenLab price: $1.88
Discount: -70%

Tool fees

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

Web search

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

	Official priceper 1M tokens	TokenLab priceper 1M tokens	Discount
Default spec	Input $5.00 / Output $25.00 / Cache read $0.50 / Cache write $6.25	Input $1.50 / Output $7.50 / Cache read $0.15 / Cache write $1.88	-70%
Prompt cache pricing
Cache read	$0.50	$0.15	-70%
Cache write	$6.25	$1.88	-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 8, 2026, 01:01 AM UTC

30-day success rate: 100.0%
30-day requests: 1K+
Last active: 16 hours ago

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

Open in Console

Open Claude Opus 4.6 in Console with a prompt ready to edit or send.

Help me try claude-opus-4-6 with a short message at /v1/messages. Show the reply, latency, and cost.

Use cases

Best for

Reasoning
Vision

Multi-step coding agents
Drive a terminal agent that plans a refactor, edits files, runs the suite, and repairs failures, with max effort reserved for the hardest tasks.
Long-running research sessions
Keep a session alive across many searches and document reads while compaction trims the earlier history instead of cutting the task off.
Professional knowledge work
Draft financial models, legal summaries, and analyses where a wrong answer costs more than extra latency.

Prompt examples

Plan a migration of this Express service to Fastify, list the files affected in order, then make the first change and run the tests.

Search these three filings for revenue recognition changes between years and summarize the differences with page references.

Review this pull request for concurrency bugs, and rank the issues by how likely each is to hit production.

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 Claude Opus 4.6 good at?

Complex agentic coding, planning, and professional analysis. Anthropic's launch post reports top results on Terminal-Bench 2.0, Humanity's Last Exam, and BrowseComp, and early testers describe it breaking large tasks into steps and finishing them rather than stopping halfway.

How is Opus 4.6 different from Opus 4.7?

Opus 4.7 adds higher-resolution image input, an xhigh effort level, task budgets, and a new tokenizer, and follows instructions more literally. Opus 4.6 predates those changes, so prompts that relied on loose interpretation may behave differently when you switch.

Does Claude Opus 4.6 support extended thinking?

It supports adaptive thinking, where the model decides how much to reason and you steer it with the effort parameter. The older manual extended-thinking mode with a token budget is marked deprecated on this model.

Is Claude Opus 4.6 still worth using?

Use it when an existing pipeline is validated on it and you cannot re-run evaluations yet. For new work Anthropic recommends migrating to Opus 5.5, which is the current Opus model and improves on it for long coding and knowledge tasks.

How much does Claude Opus 4.6 cost?

On TokenLab, Claude Opus 4.6 costs Input $1.50 / Output $7.50 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 4.6?

Claude Opus 4.6 accepts up to 1,000,000 tokens of context and returns up to 128,000 tokens in one response.

Which endpoint should Claude Opus 4.6 use?

Use https://api.tokenlab.sh/v1/messages for Claude Opus 4.6. The request example below shows the matching code shape.

Which operations does Claude Opus 4.6 support?

Claude Opus 4.6 supports Text to text. Select an operation above to see its endpoint and request example.