Costs
Claude Code's pricing is hardcoded in the source. Knowing the numbers helps you make informed decisions, especially about fast mode, which costs 6x more for the same model.
4 model tiers 6× fast mode premium $0.01 per web search
! /fast costs 6x more for the same model
Fast mode does NOT switch to a different model. It runs the same Opus 4.6 with higher
priority throughput. The input cost jumps from $5/Mtok to $30/Mtok, a 6x premium. Only
use it when response speed is genuinely worth the price difference.
Pricing per model
| Model | Input /Mtok | Output /Mtok | Cache Read | Cache Write |
|---|---|---|---|---|
| Haiku 4.5 | $1 | $5 | $0.10 | $1.25 |
| Sonnet 4.x | $3 | $15 | $0.30 | $3.75 |
| Opus 4.5 / 4.6 | $5 | $25 | $0.50 | $6.25 |
| Opus 4.6 (fast mode) 6× premium | $30 | $150 | $3 | $37.50 |
| Web Search | $0.01 / request | |||
Mtok = million tokens
Default model selection
Max / Team Premium subscribers
Opus 4.6 [1m]
1M context window, most capable model
All other users
Sonnet 4.6
200K context window, strong performance
i Cache reads are 90% cheaper than fresh reads
The prompt cache dramatically reduces costs for the static portion of the system prompt
(everything above
__SYSTEM_PROMPT_DYNAMIC_BOUNDARY__).
Long sessions benefit from cache hits on repeated reads of large files too.