Long context

claude-sonnet-4.6-thinking API pricing

Anthropic's Claude series — long context, dependable tool calling, and the native models behind Claude Code.

Model IDanthropic/claude-sonnet-4.6-thinking

ChatStreamingTool callingVision

Get API Key Run your first request

Call summary

Input

$3.00$0.90/ 1M tokens

Cache write $1.875

Output

$15.00$4.50/ 1M tokens

Cache read $0.15

Context Window

1.05M

API Endpoints

Available Configurations

claude-sonnet-4.6-thinking API endpoints

One API key — call this model through any of the 2 endpoints below.

POST/v1/chat/completionsOpenAI-compatible · Works for every model

Docs

POST/v1/messagesAnthropic-native · Native for Claude Code

Docs

Production reliability

Privacy-safe aggregates from real Router One model calls, so you can assess stability and response latency before integrating.

Successful response share

98.20%

Success 2xx

98.20%

Rate limit 429

0.90%

Server 5xx

0.90%

TPS

36.56

tokens/s

Avg. time to first token

15.5 s

Avg. latency

21.95 s

Includes production requests ending in 2xx, 429, or 5xx. Average latency uses successful requests; time to first token and TPS require complete observations from successful streams. Public thresholds are 100 requests and 5 independent principals.

Updated Sep 9, 2026, 8:59 PM UTC

More Claude models on Router One

10 other Claude models on the same gateway and API key — each page lists endpoints, posted price, and context window.

Call claude-sonnet-4.6-thinking from your code

The same API key works across all 2 endpoints below.

Endpoint

/v1/chat/completions

Model ID

anthropic/claude-sonnet-4.6-thinking

Routing mode

auto

Request example

bash

curl -X POST \
  https://api.router.one/v1/chat/completions \
  -H "Content-Type: application/json" \
  -H "Authorization: Bearer YOUR_API_KEY" \
  -d '{
    "model": "anthropic/claude-sonnet-4.6-thinking",
    "messages": [
      {"role": "user", "content": "Hello!"}
    ]
  }'

Get API Key

claude-sonnet-4.6-thinking at a glance

claude-sonnet-4.6-thinking is a Claude-series text model on Router One. claude-sonnet-4.6-thinking thinks by default: a request that omits the thinking field is sent with adaptive thinking, and a request that already carries a thinking value (disabled included) is not given a second one — the API compatibility facts state the per-generation rules. Posted rate: $0.90 in / $4.50 out per 1M tokens (was $3.00 in / $15.00 out). Context window: 1.05M tokens. Accepts text, image; returns text. Endpoints: POST /v1/chat/completions (OpenAI-compatible), POST /v1/messages (Anthropic-native, the Claude Code path).

claude-sonnet-4.6-thinking FAQ

How much does claude-sonnet-4.6-thinking cost on Router One?

claude-sonnet-4.6-thinking is billed per token on Router One: $0.90 in / $4.50 out per 1M tokens, pay as you go.

Which endpoints serve claude-sonnet-4.6-thinking?

claude-sonnet-4.6-thinking is served on 2 endpoints — POST /v1/chat/completions (OpenAI-compatible), POST /v1/messages (Anthropic-native, the Claude Code path). The same Router One API key works on each.

What is claude-sonnet-4.6-thinking's context window?

claude-sonnet-4.6-thinking accepts up to 1.05M tokens of context per request on Router One.

Start using claude-sonnet-4.6-thinking

Create an API key and call claude-sonnet-4.6-thinking at $0.90 in / $4.50 out per 1M tokens on Router One — pay as you go, with per-request cost and latency visibility.

Get API Key Plans & pay-as-you-go pricing Compare models side by side Quickstart guide