---
format: "aidr-story-markdown/v1"
id: "827ff1a721db5feb2bc28ac79f61f7e243d1ac3745ccc3505e265076f6580bc5"
canonical_url: "https://aidr.today/827ff1a7?lang=en"
title: "Gemma-4-26B-A4B with 39-69% less thinking at same accuracy"
lang: "en"
requested_lang: "en"
available_langs: ["en","vi"]
translation_fallback: null
fallback_fields: []
published_at: "2026-10-23T14:43:02.000Z"
category: "Models"
topics: ["google","inference"]
source_urls: ["https://bottlecapai.com/post/thinkingcap-gemma-4-26b/","https://news.ycombinator.com/item?id=50209907"]
summary: "The fifth model in the ThinkingCap series reduces how much Google's gemma-4-26B-A4B-it thinks, without compromising the quality of its answers."
---

# Gemma\-4\-26B\-A4B with 39\-69% less thinking at same accuracy

> [Open the canonical story](<https://aidr.today/827ff1a7?lang=en>)

**Published:** 2026-10-23T14:43:02.000Z
**Category:** Models
**Topics:** google, inference

## Summary

The fifth model in the ThinkingCap series reduces how much Google's gemma\-4\-26B\-A4B\-it thinks, without compromising the quality of its answers\.

## Sources

- [Story source](<https://bottlecapai.com/post/thinkingcap-gemma-4-26b/>)
- [Discussion](<https://news.ycombinator.com/item?id=50209907>)