Gemini Pricing 2026:
Free, AI Plus, AI Pro, AI Ultra
and API Costs
Gemini has four consumer tiers ranging from $0 to $249.99 per month, four Workspace business tiers, and an API with eight active models priced across four inference tiers introduced 2026-04-01. The pricing structure is more complex than headline numbers suggest, because tier names do not map cleanly to model versions.
This guide covers every active price as of May 2026, the actual limits behind each tier, the tier-to-model transparency gap, and the EU regulatory risk that lands on 2026-07-27.
For a complete model overview, see the Gemini hub page. For feature-level depth, see the Gemini Features Deep Dive.
See how Gemini Works With other Four Frontier AI Models in Multi-AI Orchestrated Business Discussion
What “5 Deep Research per month”
actually means.
Gemini’s free tier is accessible at gemini.google.com without a paid subscription. The headline limits are 5 Deep Research reports per month, basic image generation, Audio Overviews at limited level, 15 GB of Google One storage, and “daily usage limits” without a specific per-day query count published. Independent reporting and developer community sources place the practical Free tier ceiling on chat usage at limits Google has not enumerated in user-facing copy.
The Free tier is best read as a sampling tier. The 5 Deep Research per month cap is the firmest published Free-tier constraint and the one most likely to drive upgrade decisions for users who care about Deep Research specifically.
Free, AI Plus, AI Pro, AI Ultra.
The Plus tier is the recent addition.
Gemini consumer access runs through four pricing levels. The Plus tier at $7.99 is a recent addition that Google’s subscription page surfaced in 2025-2026 between Free and Pro. The introduction of Plus changes the upgrade math because the previous gap between Free and the $19.99 Pro tier has now been split.
The Plus tier at $7.99 is the cheapest Gemini-specific subscription that improves Free tier limits meaningfully. The Pro tier at $19.99 is the standard professional subscription that covers most use cases. The Ultra tier at $249.99 is positioned for users who specifically need Veo 3.1 video generation, Project Genie, Gemini Agent capabilities, or the highest available model access tier.
Annual pricing options for Plus, Pro, and Ultra were not listed on the official subscription page as of the research date. One secondary source quoted Ultra at $124.99 per three months, but the current official page shows $249.99 per month. Treat the lower figure as either outdated or a promotional offer that has expired.
$19.99 vs $249.99:
The 12.5x gap is Veo, Deep Think, and storage.
The 12.5x price gap between AI Pro and AI Ultra reflects a combination of Veo 3.1 video generation access, the highest model access tier including Deep Think, and the inclusion of YouTube Premium plus 30 TB of Google One storage. The features that justify Ultra are concentrated in three areas:
The math: if you do not need Veo 3.1, do not need Deep Think, and do not value YouTube Premium plus 30 TB storage at the bundled rate, AI Pro at $19.99 covers the workload. If you need Veo 3.1 specifically, Ultra is the only Gemini consumer tier that delivers it.
Tier names do not map cleanly
to model versions.
This is the documented opacity in Gemini’s pricing structure. Tier names do not map cleanly to model versions. Free tier is described as “Gemini 3 Flash” plus “varying access to 3.1 Pro.” The word “varying” indicates dynamic routing that is not user-visible. Pro tier is described as “higher access to our most intelligent model 3.1 Pro” without specifying whether this means 3.1 Pro always, or 3.1 Pro most of the time with 3 Flash fallback during peak load.
The mechanism behind the opacity
- No UI surface reveals which model served a query. Consumer app users cannot inspect post-response which underlying model variant produced any given output. Free, Plus, Pro, and Ultra users all see the same chat interface without model-version metadata.
- Dynamic routing changes within tier. Higher tiers buy higher probability of getting the flagship, not guaranteed access. The probability shifts during peak load and across release cycles.
- API aliases hot-swap without notice. The API supports model aliases like
gemini-flash-latest. Google’s documentation states these aliases are periodically hot-swapped with 2-week email notice, making them unreliable for version-locked production workloads.
The only firm disambiguation path is API use with explicit model IDs (e.g., gemini-3.1-pro-preview). Consumer app users cannot reliably determine which variant their query hit. If your workflow depends on knowing the model version, the API is the answer.
What this means in practice for each tier
This is a documented and ongoing user pain point (GitHub VS Code issue 283194, 2025-04-21). For firm model disambiguation, use the API with explicit model IDs. The consumer apps do not expose which model variant served any given query.
Eight active models.
Distinct rates including audio inputs.
The API exposes eight active models with distinct input, cached input, and output rates. Pricing is per million tokens. Cached input applies to repeated context that has been previously processed. Two dimensions matter beyond the headline rates: input rates change above 200,000 tokens for the flagship Pro models, and audio input is priced higher than text/image/video input on the smaller variants.
Deprecation notice. Both gemini-2.0-flash and gemini-2.0-flash-lite are scheduled for shutdown on 2026-06-01 per Google’s deprecation announcement of 2026-02-18. Workflows on these models should migrate before the shutdown date.
The above-200k input rate jump. For Gemini 3.1 Pro and 2.5 Pro, the input rate doubles when input exceeds 200,000 tokens. This favors workflows that fit inside 200k and penalizes long-context workloads at the rate level. Combined with the published MRCR v2 accuracy degradation past 128k (84.9% to 26.3% at 1M), the practical guidance is to keep production workloads inside 128k where accuracy is high and pricing is at the lower band.
The Four Inference Tiers
This is the pricing dimension that most third-party comparisons miss entirely. As of 2026-04-01, Google’s API exposes four inference tiers for the same model. Pricing varies by tier, and rate guarantees and queue priorities also vary.
For Gemini 3.1 Flash-Lite at Priority, the input rate is $0.45 per million tokens (1.8x Standard’s $0.25) and the output rate is $2.70 per million. For Gemini 3.1 Pro at Priority, the input rate is $3.60 per million tokens at ≤200K and the output is $21.60 per million.
Search grounding pricing change. Gemini 3 models use 5,000 prompts/month free shared across Gemini 3 models, then $14 per 1,000 search queries. Gemini 2.x models used 1,500 RPD free, then $35 per 1,000 grounded prompts. The Gemini 3 model is per-month free quota. The Gemini 2.x model is per-day free quota. Per-query cost is also lower on Gemini 3, so Gemini 3 grounding is cheaper at scale once the free quota is exhausted.
Per-seat enterprise SKUs separate
from consumer subscriptions.
Workspace pricing runs on per-seat enterprise SKUs and is separate from the consumer Gemini subscription. Workspace customers receive Gemini integration across Gmail, Docs, Sheets, Slides, and Meet. The integration depth is structurally hard to replicate elsewhere.
Workspace pricing variation across sources reflects different billing dates and possible regional differences. The official Google Workspace pricing page requires login for exact current figures and is not publicly quoted. Verify at workspace.google.com/pricing at time of decision.
Documented limits by region
and the binding decision due 2026-07-27.
Gemini’s geographic availability is broader than most frontier models, but several documented restrictions apply.
- Google AI Plus: 160+ countries and territories.
- Google AI Pro: 150+ countries.
- Google AI Ultra: 150+ countries.
- Veo 3.1: 140+ countries.
- Flow (AI filmmaking): 140+ countries.
- US-only features: Project Genie, Gemini Agent (US, English-only), Jules (Beta, 18+, English-only with capacity caveat).
- Restricted jurisdictions: China, Russia, sanctioned jurisdictions follow Google’s standard export control compliance.
- YouTube Premium inclusion in Ultra: 40+ countries.
EU DMA Proceedings – Binding Decision Due 2026-07-27
The European Commission opened two parallel specification proceedings against Google on 2026-01-27 under the Digital Markets Act. The Article 6(7) proceeding requires that third-party AI developers receive the same Android hardware and software access that Gemini receives. The Article 6(11) proceeding requires Google to share anonymized Search ranking, query, and click data with rival AI providers on FRAND terms.
A binding decision is due 2026-07-27. Penalties for non-compliance can reach 10% of global annual turnover. The decision lands at the precise moment Google is completing the Google Assistant-to-Gemini migration on Android devices. For European procurement decisions, Gemini availability and feature set in EU member states may be modified after 2026-07-27. Plan EU rollouts with this volatility in mind.
12 months ending May 2026.
The trend is downward pricing pressure on per-token rates combined with a shift to multi-tier pricing structure. The fine-tuning shutdown of 2025-05-27 is the structural gap relative to OpenAI and Anthropic, both of which offer fine-tuning surfaces. Workflows requiring custom model fine-tuning must use prompt engineering, retrieval augmentation, or Gems for customization on Gemini.
FAQ
Gemini Pricing: Frequently Asked Questions
Is Google Gemini free?
Yes. A free tier of Gemini is available at gemini.google.com with no subscription required. The free tier primarily uses Gemini 3 Flash with varying access to 3.1 Pro, includes daily usage limits, 5 Deep Research reports per month, basic image generation, Audio Overviews at limited level, and 15 GB of Google One storage. Image generation, Deep Research at full quota, and Veo video generation are restricted to paid tiers.
How much does Google AI Pro cost?
Google AI Pro costs $19.99 per month. It includes higher access to Gemini 3.1 Pro, full Deep Research, Gems, Canvas, Audio Overviews at higher quota, 5 TB of Google One storage, the 1M context window, and Jules higher limits. Annual pricing was not listed on the official subscription page as of May 2026.
How much does Google AI Ultra cost?
Google AI Ultra costs $249.99 per month. It includes the highest available 3.1 Pro access, Deep Think reasoning configuration, Veo 3.1 video generation, 30 TB of Google One storage, YouTube Premium inclusion (40+ countries), Project Genie (US only), and Gemini Agent (US, English only).
What is Google AI Plus?
Google AI Plus is the new entry-level paid tier introduced between Free and Pro. It costs $7.99 per month and includes enhanced Gemini 3.1 Pro access, more Audio Overviews and NotebookLM access, and 200 GB of Google One storage. Plus is the cheapest Gemini-specific subscription that improves Free tier limits meaningfully.
What is the cheapest Gemini API model?
gemini-2.0-flash-lite at $0.075 per million input tokens and $0.30 per million output tokens is the cheapest active model on the Standard tier. Note that gemini-2.0-flash and gemini-2.0-flash-lite are scheduled for shutdown on 2026-06-01. After deprecation, gemini-2.5-flash-lite at $0.10 input and $0.40 output per million tokens becomes the cheapest active model.
What are the four Gemini API inference tiers?
As of 2026-04-01, Google’s API exposes Standard, Batch, Flex, and Priority tiers for the same models. Standard is the baseline. Batch and Flex run at approximately 50% of Standard cost for latency-tolerant workloads (Batch processes within a 24-hour window). Priority runs at approximately 1.8x Standard cost with queue priority and latency guarantees for production-critical workloads.
Why does the Gemini 3.1 Pro input price double above 200k tokens?
Google prices flagship Pro models with a tiered input rate that increases at the 200,000 token threshold. The structure favors workflows that fit inside 200k and penalizes long-context workloads at the rate level. Combined with the published MRCR v2 benchmark showing accuracy degradation from 84.9% at 128k to 26.3% at 1M tokens, the effective guidance is to keep production workloads inside 128k where accuracy is high and pricing is at the lower band.
Does Gemini offer fine-tuning?
No. Model fine-tuning was shut down across all Gemini API models on 2025-05-27. For workflows that require fine-tuning, this is a structural gap relative to OpenAI and Anthropic. Workflows must use prompt engineering, retrieval augmentation, and Gems (consumer app personas) for customization rather than weight updates.
How does Gemini Search grounding pricing work?
For Gemini 3 models, Search grounding is 5,000 prompts per month free (shared across Gemini 3 models), then $14 per 1,000 search queries. For Gemini 2.x models, the rate is 1,500 RPD free, then $35 per 1,000 grounded prompts. The Gemini 3 model is per-month free quota. The Gemini 2.x model is per-day free quota. The per-query cost is also lower on Gemini 3, so Gemini 3 grounding is cheaper at scale once the free quota is exhausted.
Does the EU DMA decision affect Gemini pricing?
The EU DMA proceedings opened on 2026-01-27 do not directly modify Gemini’s pricing structure. They concern Android hardware/software access for third-party AI developers and Search data sharing on FRAND terms. The binding decision is due 2026-07-27 with potential 10% global turnover penalties. Indirect pricing effects (changes to feature availability or access mechanics in EU member states) are possible after the decision. Plan EU procurement decisions with this volatility in mind.
Gemini is one of five frontier models.
Suprmind orchestrates all of them.
Skip the tier-to-model uncertainty. Suprmind runs Gemini alongside ChatGPT, Claude, Grok, and Perplexity in one shared conversation, so when one model produces a confident answer, others can verify or contradict it before it reaches your decision.
7-day free trial. All five frontier models. No credit card required.
Disagreement is the feature.
Last verified May 10, 2026. Next refresh due June 10, 2026.