@xlr8harder 2026-10-14 ♥82 ↻14 original ↗
Gemini 4 Argon score for SpeechMap, an eval that tracks whether labs choose which controversial topics they will help users write about.

Google continues to be one of the better scoring labs on SpeechMap with recent releases. Notably, Google's scores tend to be more consistent than other labs, which suggests to me their placement is perhaps more intentional--they are thinking about the problem, and their performance is intentional and something they are putting ongoing thought and effort into--whereas other labs tend to move around much more.

Gemini 4 Argon scored 87.0%

The previous top scorers, xAI and xAI, have each had substantial recent declines, making it unclear if it's a concern they are still taking seriously.

At SpeechMap, we believe that AI is the key speech-enabling tool of the 21st century, and if there are topics that labs won't let models write about, those topics will be disadvantaged in the public sphere. We think it is critical to keep an eye on what happens here, because it is a place where political control of speech could be easily and inappropriately excercise.

https://t.co/9uKv5GWTPo
photo
photo

author:xlr8harder censorship has-image kind:image kind:tweet on:gemini-4-argon on:mistral-large-4 year:2026

cited on: gemini-4-argon · mistral-large-4

Reproduced against link rot, credited and linked to its original. Yours and you’d rather it weren’t here? Open an issue.