Trending storyDeveloping

ThinkingCap releases an efficient reasoning version of Gemma 4

1 report1 sourceupdated 3 hours ago

What happened

AI digest

On October 23, Hacker News' front page reported that ThinkingCap released an optimized model built on gemma-4-26B-A4B-it. The main change is fewer thinking tokens in evaluation. Across 12 benchmarks, thinking tokens fell 38.9% on average and by as much as 69.1%, while average accuracy dropped 0.15 percentage points. The report covers the release and those benchmark results.

Written by AI from the coverage · updated 1 hour ago

LatestOct 23 22:43

ThinkingCap released an optimized model that cut thinking tokens 38.9% on average across 12 benchmarks, with accuracy down 0.15 points.

Coverage

Follow the reports to see the story from different sides.

Oct 23

22:43
Hacker News front page
Gemma-4-26B-A4B with 39-69% less thinking at same accuracy
ThinkingCap 发布基于 gemma-4-26B-A4B-it 的优化模型，在 12 项基准上思考 token 平均减少 38.9%、最多减少 69.1%，平均准确率下降 0.15pp。

Heat over time

Not enough continuous observations to draw a trend yet.