Neutrality Project

We've put Gemini 3.5 Flash-Lite to the test. The model appear to be extremely censored which impact the results.

Part of our roadmap is understanding better what kind of topics are censored in order to granularly understand better the leaning.

Gemini 3.6 will come in later.
neutralityproject.org

As Google releases their new model, we felt like coming back on one of our discoveries from our first release.

What happens to Gemma's politics when you remove its guardrails? We measured it. 🧵

hmmmm...

We've just merged its first community contribution! 🚀

6 new models benchmarked and submitted via open PR: DeepSeek, Kimi K2, Hunyuan, Nova Pro, Command A, Jamba and two new countries on the map. 24 models total, from 6 countries.

And because trust is the whole point: community runs are displayed flagged as "pending re-verification" until we reproduce them ourselves. Open methodology means anyone can contribute, and nobody, including us, gets taken on faith.

Thank you bricepirard-spec. Who's next?

neutralityproject.org/results.html

When AI embeds itself in decision-making, who oversees its influence?

Right now: nobody.

We vows to build and open-source the methodologies, datasets and benchmarks that change that. Nothing stays behind closed doors.

Helps us fund the research: donate.neutralityproject.org