Anthropic Claude
Sonnet 4.6 pricing
$3/M input, $15/M output (with $0.30 prompt-cache reads) makes Sonnet ideal for enterprise copilots.
Last pricing check: Mar 13, 2026
Why teams choose this model
Scenario planning
Realistic cost examples
Numbers use Claude Sonnet 4.6 pricing
Claude for Excel analyst
Finance teams use Sonnet 4.6 with connectors to build models and memos hands-free.
Per request
$0.021
Per month
$294.00
Tokens sent
36,400,000
Project copilots
Program managers run shared projects with Sonnet 4.6’s adaptive thinking.
Per request
$0.0171
Per month
$478.80
Tokens sent
58,800,000
Document QA
Compliance/copilot flows lean on Sonnet 4.6 for faster responses than Opus.
Per request
$0.0234
Per month
$234.00
Tokens sent
30,000,000
Compare with
FAQs
Does Sonnet also offer 1M context?
Yes—Anthropic’s doc lists a 1M-token window (beta). Long-context pricing applies once you pass 200K tokens, similar to Opus.
How do caching costs work?
Prompt cache reads are $0.30/M tokens and writes $3.75/M. TokenTally treats the read price as the cache-hit rate for comparisons.
When should I choose Sonnet over Opus?
Sonnet 4.6 gives you most features (adaptive thinking, compaction, connectors) at half the price and lower latency.
Pricing sources
- https://platform.claude.com/docs/en/about-claude/models/overview
Checked Mar 13, 2026
- https://www.anthropic.com/news/claude-sonnet-4-6
Checked Mar 13, 2026