Oct 4, 2026
ManyPress

Advertisement

Artificial Intelligence

Starting in October 2026, Google will limit free and low-cost Gemini access to smaller models, reserving its most powerful versions for higher-priced subscriptions.

ManyPress

ManyPress

ManyPress Editorial

2 min readSource:The Decoder
Google Restricts Gemini Model Access Across Subscription Tiers

Key facts

  • •Starting in October 2026, free users will be restricted to the Flash-Lite model.
  • •The $4.99 per month AI Plus subscription will no longer include access to the Pro model.
  • •Access to all three models will require an AI Pro subscription at $19.99 per month or AI Ultra plans at $99.99 or $199.99 per month.
  • •The changes mirror the freemium model currently used by OpenAI for ChatGPT.
  • •The upcoming Gemini 4 Argon model is noted as being more resource-hungry.

Beginning in October 2026, Google will implement new restrictions on its Gemini AI models for personal account users. Those without a subscription will be limited to the "Flash-Lite" model, while the "Flash" and "Pro" models will be reserved for paying customers. Additionally, users on the $4.99 per month AI Plus plan will no longer have access to the Pro model, leaving them with only Flash-Lite and Flash.

By the numbers

monthly cost of AI Plus subscription$4.99
monthly cost of AI Pro subscription$19.99
lower-tier AI Ultra monthly cost$99.99
higher-tier AI Ultra monthly cost$199.99

New Subscription Requirements

Under the updated structure, access to the full range of Gemini models will require higher-tier subscriptions. Specifically, the AI Pro plan at $19.99 per month or the AI Ultra plans, priced at $99.99 or $199.99 per month, will be necessary to access all three models. This change marks a shift from the current free tier, which previously provided access to the 3.6 Flash model and varying access to the 3.1 Pro model.

Future Model Development

The adjustment to the subscription tiers may be intended to prepare for the release of the Gemini 4 Argon model. This upcoming model is expected to be more resource-intensive and potentially more expensive to operate than current versions.

Advertisement

This article was independently rewritten by ManyPress editorial AI from reporting originally published by The Decoder.

Artificial Intelligence