BillHsiehM

Cache hit rate

92%

Input served from cache

Higher than 10% of developers

Cache reuse

18.8×

Reads per cached token

Higher than 36% of developers

Cost per 1M tokens

$1.10

Blended cost across all tokens

Lower than 27% of developers

Output share

1%

Tokens generated as output

Higher than 99% of developers

Parallel tools

15%

Days with 2+ tools · 1.2 avg/day

Parallel machines

0%

Days with 2+ machines · 1.0 avg, 1 max

Models per active day

1.5

Distinct models used each day

Daily cost variation

2.3

Coefficient of variation; lower is steadier

Active days in span

59%

First to last active day