# Gemini-3.1-flash-lite: implicit cache stopped applying cost savings (~June 30 2026)

**URL:** <https://discuss.ai.google.dev/t/gemini-3-1-flash-lite-implicit-cache-stopped-applying-cost-savings-june-30-2026/173779>\
**Category:** Gemini API\
**Tags:** bug\
**Created:** [July 6, 2026, 3:38am UTC](https://discuss.ai.google.dev/t/gemini-3-1-flash-lite-implicit-cache-stopped-applying-cost-savings-june-30-2026/173779 "2026-07-06T03:38:54Z")\
**Posts on this page:** 3\
**Page:** 1

<div class="post-metadata">

**Author:** ![IQkamikadze](https://d1dlmcr85iqnpo.cloudfront.net/user_avatar/discuss.ai.google.dev/iqkamikadze/32/135027_2.png) [@IQkamikadze](https://discuss.ai.google.dev/u/IQkamikadze)\
**Post date:** [July 6, 2026, 3:38am UTC](https://discuss.ai.google.dev/t/gemini-3-1-flash-lite-implicit-cache-stopped-applying-cost-savings-june-30-2026/173779/1 "2026-07-06T03:38:54Z")

</div>

Since ~June 30, 2026, implicit context caching stopped applying its cost  
discount for gemini-3.1-flash-lite on my project.

- Before ~June 30 my effective input rate was ~$0.16 / 1M tokens (implicit  
cache applying the ~90% discount on my stable ~50K-token system prefix).
- Since June 30 billing shows input at the full standard rate ($0.25 / 1M),  
no “cached input” SKU, and Savings = $0.00 — even though API responses  
still report cached\_content\_token\_count \> 0 (~16%) in usage\_metadata.
- My code and request pattern have NOT changed.

Questions:

1. Is this a known regression with implicit caching for Gemini 3.x?
2. Why does usage\_metadata report cached tokens while billing applies no  
cache discount (Savings $0.00)?
3. What is the recommended way to restore cache savings for  
gemini-3.1-flash-lite — is explicit caching (caches.create) the  
supported path?

---

<div class="post-metadata">

**Author:** ![Sai\_Deepika\_K](https://d1dlmcr85iqnpo.cloudfront.net/user_avatar/discuss.ai.google.dev/sai_deepika_k/32/34803_2.png) [@Sai\_Deepika\_K](https://discuss.ai.google.dev/u/Sai_Deepika_K)\
**Post date:** [July 14, 2026, 12:01pm UTC](https://discuss.ai.google.dev/t/gemini-3-1-flash-lite-implicit-cache-stopped-applying-cost-savings-june-30-2026/173779/2 "2026-07-14T12:01:31Z")

</div>

Hello @IQkamikadze,

I have DM’ed you for details.

---

<div class="post-metadata">

**Author:** ![Mathew](https://d1dlmcr85iqnpo.cloudfront.net/user_avatar/discuss.ai.google.dev/mathew/32/32078_2.png) [@Mathew](https://discuss.ai.google.dev/u/Mathew)\
**Post date:** [August 24, 2026, 7:54pm UTC](https://discuss.ai.google.dev/t/gemini-3-1-flash-lite-implicit-cache-stopped-applying-cost-savings-june-30-2026/173779/3 "2026-08-24T19:54:09Z")

</div>

Hi I made a post regarding this issue which seemed to happen around the time when Gemini 3.6 flash was released. I found that the minimum requirement for the caching to kick in is around 100k and it ONLY cache the first ~80k! I just checked today and it’s still not working as it used to.

This is totally not acceptable from Google if they wanted companies to adopt gemini for commercial productions. If my app was live, I could be losing money thanks to this issue with implicit caching because it would have increased my cost by 100% even with all of the mitigation to deal with their current caching behavior.

Anyone else experiencing this issue?

> [@Can't get the implicit caching to work](https://discuss.ai.google.dev/t/cant-get-the-implicit-caching-to-work/97205/12):
>
> Hi, I noticed implicit caching behavior changed after gemini 3.6 flash was released. I used to get \>90% cache rate but after the release, the cache rate fluctuated between 0% to ~65% for my app. I did a thorough investigation using lots of claude quota and lot of diagnostic data from my chat history. I collect the in/out token and cache token count as diagnostic on each turn including the data for each tool call made by the main chat AI. Here’s what I learned based on diagnostic logs from a cha…