elevated Gemini Ai Ultra Pricing · Research

Gemini App Limits Cut: Free Users Get Flash-Lite Only, and AI Plus Loses Gemini Pro

Data graphic: dot plot of Gemini usage multipliers by plan, 1x with no plan, 2x AI Plus, 4x AI Pro
AK

Threat intelligence editor · Published Oct 14, 2026, 12:52 PM EDT

From 9 October Google limits free Gemini users to Flash-Lite and takes Pro from AI Plus. Here is what each plan keeps, and where Deep Think is still unclear.

Google restricted the Gemini app for users without a paid plan to its smallest model, Flash-Lite, from 9 October 2026. AI Plus subscribers lose access to Gemini Pro on a date Google emails to each account. AI Pro and AI Ultra keep every model they had. The change applies to personal accounts; Google's support page does not cover work or school accounts.

What changed, by plan

Prices are US figures reported by 9to5Google. Google's help page lists none. The "Before" column is inferred from what each group loses, because Google publishes no earlier grid. One secondary source says free-tier Pro access may have varied.

PlanUS price (9to5Google)BeforeAfter 9 OctUsage multiplier
No planFreeFlash-Lite, Flash (3.6), Pro (3.1)Flash-Lite (3.5) onlyStandard (1x)
AI Plus$4.99/moFlash-Lite, Flash, ProFlash-Lite, Flash; no Pro2x
AI Pro$19.99/moFlash-Lite, Flash, ProUnchanged; Deep Think disputed, see below4x
AI Ultra$99.99 or $199.99/moFlash-Lite, Flash, Pro, Deep ThinkUnchanged5x or 20x higher than AI Pro

Free users on Auto see most prompts routed to Flash-Lite. With smart model selection off, every response uses Flash-Lite. The sources do not say which Ultra price carries the 5x tier and which the 20x tier.

Data graphic: Gemini model access after 9 October for Free, AI Plus and AI Pro, with US prices

Model access after 9 October by plan. Prices are US figures from 9to5Google; the before state is inferred.

Usage limits and effort levels

Limits have been compute-based since 17 May 2026 (24 July for under-18s). They refresh every five hours until a weekly cap is reached. Google publishes no numeric caps. Usage depends on prompt complexity, features and chat length. Media generation, Deep Research, the Pro model, extended thinking and Deep Think drain the allowance faster.

The multipliers are Google's: standard for no plan, 2x for AI Plus, 4x for AI Pro, and "5x or 20x higher than AI Pro" for Ultra. Because the Ultra figures are stated against AI Pro, not against the standard allowance, this article charts only the first three plans and does not convert Ultra to a standard-plan multiple.

October also brings low, medium and high effort levels for each available model, replacing the single "Extended thinking" toggle. Higher effort improves thoroughness and uses more of the limit. 9to5Google says the app "looks to be adding" them, with no day given.

Deep Think and AI Pro: unresolved

9to5Google says Deep Think, previously on the $99.99 and $199.99 plans, comes to AI Pro for "maximum parallel reasoning." Google's help page first said so too. By 6 October the US-English version reportedly named only AI Ultra, while other language versions still listed AI Pro. Treat Deep Think on AI Pro as unconfirmed until it shows in your account. For that reason it appears in neither image here.

Gemini 4 Argon arrives first for AI Ultra. Where it fits relative to Pro is unclear.

What teams should do

  • Check which Gemini plan staff use for work. Prompts that depend on Pro reasoning will fall back to Flash-Lite or Flash on free and Plus accounts.
  • Do not upgrade to AI Pro for Deep Think until the feature appears in an account.
  • Expect heavy use of high effort, Deep Research and media tools to hit the five-hour and weekly limits sooner.
  • Use workspace or API plans for production work. The consumer app limits are not a service guarantee.
  • For how the Ultra tiers compare with rival premium plans, see Google AI Ultra vs Claude Max vs ChatGPT Pro for developers.

Sources

Keep reading

All latest →
  1. watchResearchLiquid LFM2.5-24B-A2B vs Qwen3.6-35B-A3B: Better at Agents, Weaker at Coding4 min
  2. watchResearchReflection Beam: A 501B Open-Weight MoE Claims 80.9 on SWE-bench Verified, but the Weights Have Not Shipped3 min
  3. stableResearchAI Floor Plan Generator Free Tiers Compared: Editable Plans and Code Checks8 min
  4. stableResearchUnderdog Saluki 27B Squeezes Qwen3.8-27B Into a 7.89 GB 2-Bit GGUF for 16 GB GPUs5 min
  5. stableResearchAleph Alpha's Kolibri-1 Puts a 78B Open-Weight MoE With 1M-Token Context on a Single H2005 min
  6. watchResearchGemini API Migration Checklist: 3.7 and 3.5 Flash Deprecated, Deep Research Preview Ends Oct 234 min