Evaluation firm ValsAI's RSI Index rankings show Google's Gemini 4 Neon at No. 4 on the index, making it Google's best-performing model and 3.8 points ahead of the previous generation's best, Gemini 4 Argon. The rankings also show that Google's acceleration over the past two months has put it firmly ahead of OpenAI on the RSI Index.

Key ranking data

According to the RSI Index published by ValsAI, Gemini 4 Neon ranks fourth, Google's highest placement on the index. Its score is 3.8 points higher than Gemini 4 Argon, Google's previous best model. The findings indicate that Google's progress over the past two months has firmly pushed it past OpenAI on the RSI Index.

ModelRSI Index rankScore change
Gemini 4 NeonNo. 4Up 3.8 points from Gemini 4 Argon
Gemini 4 ArgonGoogle's previous bestBaseline

Visualizing effort tiers

Around the time the rankings were released, Simon Willison once again used the classic "pelican riding a bicycle" test to visually compare output differences across effort tiers within the same model family. The models compared included Gemini 4 Carbon, the new Gemini 4 Neon and Gemini 4 Hydrogen, each generating SVG at low, medium and high effort.

Hugo Bowne-Anderson shared the test, calling it an interesting and effective way to compare model effort tiers visually — the differences in style and detail from the same prompt at different effort levels are immediately clear.

Reactions

KOL teortaxesTex commented that Anthropic still leads in internal capability, but that Google is at least a notch ahead of Neon/Carbon.

This is an area that "can't be killed" — companies with this kind of capability and resources never hit a wall; more investment means faster progress.

What to watch

Future updates to the RSI Index will reflect ranking shifts among Google, OpenAI, Anthropic and other vendors. The actual performance differences across Gemini 4's effort tiers could also influence developers' choice of model versions.