Neon Field GuideIndependent reference

Editorial Policy

How Neon Field Guide checks sources, labels benchmarks and maintains its Gemini 4 Neon coverage.

Which sources take priority?

We use Google’s launch announcement to establish what was announced, Google AI Studio’s rate card for developer API pricing and DeepMind’s model cards for specifications and model-card evaluations. When two Google products list different limits, we name the product and context instead of combining them. A launch announcement can document the initial rollout while later documentation reflects the current terms. A meaningful claim should have a source the reader can open.

How we handle benchmark comparisons

We record the model, benchmark and any evaluation version or harness reported in the source. We label provider-reported results as such. If two cards use different benchmark versions or clearly different scoring methods, we call that out and do not calculate a gap between them. Even where labels match, settings and dates may differ. The browser-based calculator, worked examples and worksheet are separate from benchmark results. Worked examples use stated inputs and are not measurements from an actual customer account.

When we review and update pages

We review relevant pages when a provider publishes a new model card, a pricing change, an access change or a correction that affects the guide. We also review reader reports as they arrive. Review is event-driven, and publication does not follow a daily schedule. When the source does not give enough detail, we state what we could verify and provide the relevant official link. Source review dates describe when a claim was checked.

Corrections and independence

Send a correction through the contact page. Give the affected page, the disputed text and any source that supports the change. We compare the report with the original evidence before updating published figures or guidance. Google does not operate this site. Analytics and publisher verification are site operations and do not determine which benchmark rows or product limitations we discuss.