gemini-4IMAGE
Gemini 4 Argon prompt cache hits only 73%, costing devs up to 1.48x
Claude and GPT hit the prompt cache every turn (228/228) while Gemini 4 Argon hit only 167/228, and the 27% miss rate made agent loops cost up to 1.48x a perfect cache.
NZ ☄️
@CodeByNZArchived 19 days agoOriginally posted 2026.10.17
NZ ☄️ @CodeByNZOriginally posted 2026.10.17
🚨Google's Gemini cache is costing developers extra. Claude and GPT hit the prompt cache on every turn (228 of 228). Gemini 4 Argon only hit 73% of the time (167 of 228), so it paid up to 1.48x what a perfect cache costs. A 27% cache miss rate in agent loops. Insane.
x.com/codebynz/status/2111372736962687068Archived 2026.10.18 10:09