Google shipped its fourth Gemini 4 model in 7 weeks.

Neon ranks third on the Artificial Analysis index, behind Claude Fable 5.2 and GPT-6.2 Astra.

Neon costs $3 per million input tokens and $15 per million output, and Glean says it runs at 60% of Carbon’s cost.

One developer found Neon cost about $3.50 against $7.20 for Claude Opus 5.5 but judged Opus’s output clearly better.

Google published a 13 November paper on self-improvement memory, a method for continual recursive self-improvement.

How each outlet framed it
Zhidx
sets Neon's Artificial Analysis score (61, third) against the Carbon and Hydrogen tier split, and the 95% cache discount

Sources: Zhidx