Anthropic: Claude Sonnet 4.6

Claude workhorse for coding agents, careful analysis, and production cost control

claude-sonnet-4-6

AvailableAnthropicChatCached input 90% less

Input / Output-70%: $3.00 / $15.00$0.90 / $4.50
Context: 1M
Released: Feb 17, 2026
Max output: 128K
Modalities: VisionChat
Capabilities: Tool usePrompt CacheReasoning

About Claude Sonnet 4.6

Claude Sonnet 4.6 is Anthropic's mid-tier Claude model, released in February 2026 for coding, agents, and everyday knowledge work. It follows instructions more consistently than Sonnet 4.5 and operates a computer through screenshots, mouse, and keyboard nearly as well as Opus 4.6, while staying in the Sonnet price class. It accepts text and images and supports tool use, prompt caching, and extended thinking.

Where it works well

Coding work that spans many files: Anthropic reports that early-access developers preferred it over Sonnet 4.5 by a wide margin.
Computer use: it scores within a fraction of a point of Opus 4.6 on OSWorld-Verified, Anthropic's reported benchmark for operating desktop software.
Long agent runs that call many tools, where Anthropic reports a clear gain over Sonnet 4.5 on its MCP-Atlas evaluation.
Extended thinking can be turned up for hard problems and left off for fast, cheap turns.

When to choose another model

Opus-class Claude models remain the better choice for the hardest multi-step reasoning and the longest autonomous tasks.
For classification, extraction, and other high-volume simple calls, Claude Haiku 4.5 answers faster at a lower rate.
It writes text only; image, audio, and video output need a separate model.

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-sonnet-4-6",
    "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 $3.00 / Output $15.00 / Cache read $0.30 / Cache write $3.75
Verified price: Input $0.90 / Output $4.50 / Cache read $0.09 / Cache write $1.13
Discount: -70%

Prompt cache pricing

Cache read

Official price: $0.30
Verified price: $0.09
Discount: -70%

Cache write

Official price: $3.75
Verified price: $1.13
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 $3.00 / Output $15.00 / Cache read $0.30 / Cache write $3.75	Input $0.90 / Output $4.50 / Cache read $0.09 / Cache write $1.13	-70%
Prompt cache pricing
Cache read	$0.30	$0.09	-70%
Cache write	$3.75	$1.13	-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 10, 2026, 02:24 AM UTC

30-day success rate: 100.0%
30-day requests: 100+
Last active: 16 hours ago

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

Open in Console

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

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

Use cases

Best for

Reasoning
Vision

Coding agents
Run it behind an editor or terminal agent to plan a change, edit several files, run the tests, and fix what failed, with fewer dropped instructions than earlier Sonnet versions.
Computer-use automation
Let it fill web forms, move data between desktop apps, and confirm the outcome from screenshots when the target system has no API.
Document and contract review
Load long reports, contracts, or research papers, ask for discrepancies and risks, and get answers that cite the passages they rely on.
Support and operations assistants
Answer customer questions from a knowledge base, call internal tools to look up orders or accounts, and hand off cleanly when a human is needed.

Prompt examples

Below are a failing test and the three files it touches. Find the cause, propose the smallest fix, and explain what else the change could break.

Read this 40-page vendor contract and list every clause that limits our right to terminate, with the section number and a one-line reason for each.

You have tools to search orders and issue refunds. A customer says order 4417 arrived damaged. Decide what to check first, then act.

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 Sonnet 4.6 best at?

Coding, agent workflows, and computer use. Anthropic positions it as the model most teams should default to: it handles multi-file code changes and long tool-calling runs well, and it operates desktop software from screenshots almost as accurately as Opus 4.6. It is also a solid general model for analysis, writing, and document review.

How is Claude Sonnet 4.6 different from Claude Opus 4.6?

Sonnet 4.6 is the faster, lower-cost tier; Opus 4.6 is the more capable one. Sonnet comes close to Opus on coding and computer-use benchmarks, so most production workloads start on Sonnet. Move a task to Opus when it needs deeper reasoning over many steps or keeps failing on Sonnet.

Does Claude Sonnet 4.6 support extended thinking?

Yes. You can let the model reason before it answers and control how much effort it spends. Leave thinking off for short, latency-sensitive turns, and raise it for debugging, planning, or analysis where a wrong answer costs more than a slower one.

Can Claude Sonnet 4.6 read images?

Yes. It accepts images alongside text, so it can read screenshots, charts, scanned pages, and UI mock-ups. The same vision input is what lets it operate a computer: it looks at the screen, decides where to click or type, and confirms the outcome.

Should I upgrade from Claude Sonnet 4.5 to Sonnet 4.6?

For coding and agent workloads, yes. Anthropic reports better instruction following, less overengineering, and a large gain in computer use over Sonnet 4.5. Prompts written for 4.5 generally carry over; re-run your own evaluation set before switching production traffic.

How much does Claude Sonnet 4.6 cost?

On TokenLab, Claude Sonnet 4.6 costs Input $0.90 / Output $4.50 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 Sonnet 4.6?

Claude Sonnet 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 Sonnet 4.6 use?

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

Which operations does Claude Sonnet 4.6 support?

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