Gemini 3. 1 is a downgrade - muted and shorter answers

Gemini API
1

It feels like Gemini 3. 1 is a downgrade.

The first obvious issue is context window. You really have to coax it to give you long and complete answers. Previously you’d be able to provide a long prompt with lots of details and Gemini would incorporate all of those details into an answer.

Now it seems Gemini ingests your prompt and reduces its response to major points and drops a lot of the details. Previously, I could ask it to take multiple factors into account. Now it seems to literally drop large chunks and intent, and meaning.

I ran some prompts side-by-side in 3.0 and 3.1. 3.0 addresses the full prompt and provides a more reason. Thorough response for complex prompts with multidimensional factors.

2 Likes
2

Modify your prompt asking it to elaborate on all points.

It’s helpful that it doesn’t waste tokens if not necessary.

3

That doesn’t help. No amount of coaxing produces the coherence of 3.0.

1 Like