Claude Sonnet 3.5
Pricing & Specs

Anthropic · $3 input / $15 output per 1M tokens · 200K context window

No longer offered on OpenMark

$3
per 1M input tokens
$15
per 1M output tokens
200K
context window (tokens)

Claude Sonnet 3.5 is no longer offered on OpenMark (deprecated or superseded upstream). Its specs are kept here for reference — if you're still running it in production, it's worth benchmarking current alternatives: newer models are often cheaper AND better on the same task.

Claude Sonnet 3.5 API Pricing

TokensPrice
Input$3 / 1M tokens
Output$15 / 1M tokens
Cached input$0.3 / 1M tokens

50% discount available for batch API requests. Batch API pricing is available for this model at 50%. Cache writes and cache hits or refreshes are billed separately.

What that means in practice

1x One typical task (1K input + 500 output tokens): $0.0105
1K 1,000 runs of that task: $10.50
💡 Real cost depends on how verbose the model is on YOUR prompts — benchmark to measure actual tokens, not estimates.

Claude Sonnet 3.5 Specs

SpecValue
Context window200K tokens
Max output tokens8.192K tokens
Input modalitiestext, image
Output modalitiestext
Latency (measured by OpenMark)~1.8s median response
Reasoning modelNo
Tool / function callingYes
JSON modeYes
StreamingYes
Prompt cachingYes
Batch APIYes

Capabilities

text generation vision tool use structured outputs streaming

Notes

Model will be retired on October 22, 2025 at 9AM PT and is no longer supported by Anthropic.

FAQ

How much does Claude Sonnet 3.5 cost?

Claude Sonnet 3.5 costs $3 per 1M input tokens and $15 per 1M output tokens ($0.3 per 1M cached input tokens).

What is Claude Sonnet 3.5's context window?

Claude Sonnet 3.5 has a 200K-token context window. Maximum output is 8.192K tokens.

Can I test Claude Sonnet 3.5 on my own task?

Claude Sonnet 3.5 is no longer offered on OpenMark, but you can benchmark its current alternatives from Anthropic and other providers on your own task — free tier available.

Still using Claude Sonnet 3.5?

It's been superseded. Benchmark its current alternatives on your actual
workload — you'll likely find something cheaper AND better. Free tier available.

Test Current Alternatives — Free →