Google tightens Gemini limits amid developer pushback

Google revises Gemini usage caps after I/O backlash

Google announced new Gemini limits on Monday, two days after developers complained about hitting caps too quickly at I/O 2026. The change replaces the compute‑based throttling introduced at the conference with a tiered prompt and image quota.

Free accounts now receive five Gemini 2.5 Pro prompts per day, while the AI Pro plan grants 100 prompts and the AI Ultra plan 500. The same free tier also gets five Deep Research reports and 100 generated images daily. Upgrading to Pro or Ultra bumps the image quota to 1,000 per day. The numbers come directly from Google’s updated Help Center article.

Why the limits matter for developers

Prompt caps translate to real‑world friction for engineers who rely on Gemini for quick code snippets, documentation drafts, or research summaries. Five prompts a day barely covers casual curiosity; even the AI Pro tier’s 100 prompts can be exhausted on a busy sprint.

The image caps are equally restrictive for teams that use Gemini to generate UI mock‑ups or diagrams. A developer who needs to iterate on visual concepts may hit the 100‑image ceiling before the day ends, forcing a costly upgrade or a switch to another service.

Gemini Code Assist: a generous free tier that raises questions

In parallel, Google lifted the price tag on Gemini Code Assist for individual developers. The tool now offers 180,000 code completions per month at no cost, a figure the company says outpaces Microsoft’s GitHub Copilot, which caps at 2,000 completions and 50 chat messages per month.

The free Code Assist tier runs on the same Gemini 2.0 model used in the enterprise version. It supports a 128,000‑token context window, which should accommodate large codebases. However, the preview omits enterprise‑only features such as productivity metrics and custom AI responses. Google notes the functionality may change, leaving developers uncertain about long‑term availability.

From vague caps to concrete numbers: a shift in transparency

Before the recent Help Center update, Gemini’s limits were described in nebulous terms like “limited access” or “we may at times have to cap the number of prompts.” Those phrases gave Google room to throttle usage without clear justification, a practice that annoyed power users.

The new tiered system replaces the compute‑based limits introduced at I/O with explicit daily quotas. While the move improves predictability, it also surfaces a tension: Google must balance resource costs against a developer community that expects near‑unlimited access, especially after the company positioned Gemini as a flagship AI offering.

Technical mechanics behind the caps

Google’s compute‑based limits tied usage to the amount of backend processing each request consumed. By switching to a prompt‑count model, the company simplifies billing but obscures the underlying cost structure. A five‑prompt free tier could still consume significant GPU time if each request uses the full 128,000‑token window.

Code Assist’s 180,000‑completion allowance is framed as “enough that even prolific professional developers won’t run out.” The claim rests on an assumption about average completion length and frequency, which varies widely across projects. Without public data on average token consumption per completion, the figure remains a rough estimate.

Competitive landscape: Google versus Microsoft and OpenAI

Google’s expanded free Code Assist limits directly challenge Microsoft’s Copilot, which restricts users to 2,000 completions per month. The disparity is stark, but Copilot’s tighter caps are partly a product of its pricing model and the broader Azure ecosystem.

OpenAI’s ChatGPT and Claude from Anthropic also offer developer‑focused APIs, but they typically charge per token, making direct quota comparisons difficult. Google’s quota‑based approach could attract developers who prefer predictable limits over per‑use billing, provided the caps stay generous.

What to watch next

The next data point to track is whether Google revises the free‑tier prompt quota in response to continued developer feedback. A further increase would signal a willingness to sacrifice short‑term revenue for ecosystem lock‑in. Conversely, a rollback could indicate that the current limits already strain Google’s compute budget.

Watch for announcements about the eventual pricing of Code Assist’s preview features. If Google introduces a paid tier that restores enterprise‑only metrics, developers will need to weigh the cost against the convenience of a single‑vendor AI stack.

Finally, monitor any updates to the compute‑based throttling policy. If Google re‑introduces hidden caps under the guise of “fair use,” the community’s trust in the quota system could erode, prompting a shift toward competing platforms.

Updates

• 2026-08-04 — EON wants to move the data superhighway from ocean fiber to space lasers (source)