Google’s AI subscription lineup has a pretty awkward gap in it. You’ve got AI Pro at $20 a month, which covers most casual users just fine. Then there’s AI Ultra at $250 a month, which is a massive jump. There’s not much in between for power users who need more than Pro but won’t pay $3,000 a year. A new Google AI Ultra Lite plan might be about to change that.
Code spotted by 9to5Google’s APK Insight team in a recent Gemini macOS app update points to the new tier. The plan carries the codename “Neon” internally, and the name “Google AI Ultra Lite” is a tentative placeholder that could change before any official launch.
What We Know (And What We Don’t)
No pricing has been confirmed, but 9to5Google estimates somewhere between $50 and $150 would make sense given the existing tiers. Anthropic already has a $100 Max tier, and OpenAI added its own $100 mid-tier plan last month. Google may be following the same playbook. In terms of what Google AI Ultra Lite would actually offer, the only detail the code suggests is increased usage limits over Pro. Nothing points to new features beyond what the Ultra plan already offers. It’s more headroom for the kind of heavy usage that bumps up against Pro’s caps.
That usage ceiling is a real pain point right now. AI agents and coding tools have driven demand up sharply, and users across platforms have been hitting their limits hard. A middle tier gives Google a way to catch subscribers ready to spend more, without asking for a $250 monthly commitment.
Alongside the Google AI Ultra Lite discovery, 9to5Google also spotted plans for a dedicated usage dashboard at gemini.google.com/usage. The page would let subscribers track their remaining token budget in real time. Right now, usage limits on Gemini are pretty opaque. You often don’t know you’re close to hitting the wall until you’re already there.
If you’ve been eyeing the Ultra plan’s AI Inbox and advanced Gemini perks but can’t stomach the price, this could be the tier that bridges the gap. Google hasn’t announced anything officially, and features found in app code don’t always make it to launch.