Quick answer: Google Gemini is free to start. Paid consumer plans run $19.99/month (Pro) or $124.99 for 3 months (Ultra). API access starts at $0.10/1M tokens (Flash-Lite) up to $4.00/1M tokens (Gemini 3 Pro, large context). Pricing changes frequently — check Google's pricing page for the latest.
Google Gemini is Google's family of generative AI models capable of processing and generating text, images, code, audio, and video. It spans use cases from on-device AI (Gemini Nano) to enterprise-scale cloud inference (Gemini Pro, Gemini Ultra), and powers products from the Gemini consumer app to the Vertex AI developer platform.
Because Gemini is natively integrated into Google Search, Google Workspace, Android, and Chrome, it reaches a broader deployment surface than most AI platforms. For developers and enterprises, that integration also means Gemini API costs can scale quickly — and unpredictably — as usage grows across teams and products.
Prices listed are for the United States. Regional pricing varies.
The Free plan includes access to the Gemini app with Gemini 2.5 Flash and limited 2.5 Pro access, plus Deep Research, Gemini Live, Canvas, and Gems. Users receive 100 monthly AI credits for video generation via Flow and Whisk, basic image/video creation in Whisk, limited Veo 3 access, NotebookLM, and 15 GB of shared Google storage.
Pro expands access to Gemini 2.5 Pro with Deep Research, video generation via Veo 3.1, and 1,000 monthly AI credits. U.S. subscribers also get Gemini 3 model access. Additional benefits include Gemini in Gmail and Docs, higher limits in Gemini Code Assist and Gemini CLI, the async coding agent Jules, and upgraded NotebookLM (5× audio overviews).
Ultra is Google's most capable consumer tier, unlocking Gemini 3.1 Pro, Gemini 2.5 Deep Think, Veo 3.1, and 25,000 monthly AI credits. U.S. subscribers get Gemini 3 Pro in AI Mode for Google Search, the highest limits in Gemini Code Assist and CLI, $100/month in Google Cloud credits, YouTube Premium, and the Google Home Premium Advanced plan.
For teams building on the Gemini API or Vertex AI, total cost is shaped by several variables that compound at scale.
Model tier is the single biggest driver. Gemini 3 Pro costs 20× more per input token than Flash-Lite. Choosing the wrong model for routine tasks is the most common source of unnecessary spend.
Context window size creates cost cliffs. For 2.5 Pro and 3 Pro, input pricing roughly doubles when prompts exceed 200,000 tokens. Retrieval-augmented pipelines that include large documents can silently push every request into the higher bracket.
Reasoning tokens are billed alongside visible output. Complex prompts that trigger extended internal reasoning — even from short inputs — can inflate output costs significantly.
Batch vs. interactive mode is a 50% cost split. Batch mode carries no SLA and up to 24-hour latency, but cuts every billable request in half. This is the highest-leverage single toggle available to most teams.
Context caching lets teams pay once for large, reusable context (policy documents, codebases, catalogs) and get discounted rates on subsequent calls. Storage charges apply per hour.
Search grounding is free up to 1,500–5,000 requests per day, depending on the model, then billed at $14–$35 per 1,000 calls. Grounding costs are easy to overlook in cost forecasts.
Media generation (images via Imagen, video via Veo) uses completely separate meters — per image or per second of video — and must be budgeted independently from token-based workloads.
Gemini 3 is optimized for agentic use cases, complex multimodal reasoning, and high-volume throughput.
Gemini 3 Pricing (Per 1M Tokens, USD)
|
Model |
Input |
Output |
Context Caching |
Grounding |
Batch |
|
3 Pro |
$2.00 (≤200k) / $4.00 (>200k) |
$12.00 (≤200k) / $18.00 (>200k) |
$0.20–$0.40 + $4.50/hr |
5,000/mo free → $14/1k |
50% off |
|
3 Flash |
$0.50 (text/img/video) / $1.00 (audio) |
$3.00 |
$0.05–$0.10 + $1/hr |
5,000/mo free → $14/1k |
50% off |
|
3 Pro Image (Preview) |
$2.00 (text/img) |
$12.00 (text) / $120.00 (images) |
Same as 3 Pro |
Same as 3 Pro |
50% off |
Gemini 2.5 Pricing (Per 1M Tokens, USD)
Gemini 2.5 includes several model variants optimized for different workloads—ranging from high-performance coding and reasoning tasks in Gemini 2.5 Pro to lightweight, scalable processing in Flash-Lite. Each variant has distinct pricing tiers based on input/output token volume, media type, and usage mode (standard vs. batch). Below is a breakdown of the key pricing metrics.
|
Model |
Input |
Output |
Context Caching |
Grounding |
Batch |
|
2.5 Pro |
$1.25 (≤200k) / $2.50 (>200k) |
$10.00 (≤200k) / $15.00 (>200k) |
$0.125–$0.25 + $4.50/hr |
1,500 RPD free → $35/1k |
50% off |
|
2.5 Flash |
$0.30 (text/img/video) / $1.00 (audio) |
$2.50 |
$0.03–$0.10 + $1/hr |
Shared 1,500 RPD free → $35/1k |
50% off |
|
2.5 Flash-Lite |
$0.10 (text/img/video) / $0.30 (audio) |
$0.40 |
$0.01–$0.03 + $1/hr |
Shared 1,500 RPD free → $35/1k |
50% off |
|
Flash Native Audio (Live API) |
$0.50 (text) / $3.00 (audio/video) |
$2.00 (text) / $12.00 (audio) |
Not available |
Not available |
N/A |
|
Model |
Fast |
Standard |
Ultra |
|
Imagen 4 |
$0.02/img |
$0.04/img |
$0.06/img |
|
Imagen 3 |
– |
$0.03/img |
– |
|
Model |
Fast |
Standard |
|
Veo 3.1 |
$0.15/sec |
$0.40/sec |
|
Veo 3 |
$0.15/sec |
$0.40/sec |
|
Veo 2 |
- |
$0.35/sec |
| Service | Price | Notes |
|---|---|---|
| Gemini Embedding | $0.15/1M tokens | Text indexing & retrieval |
| 2.5 Computer Use (Preview) | $1.25–$2.50/1M input | Browser automation agents |
| Gemma 3 / 3n | Free | Open-weight, local models |
| Google Search Grounding | 1,500 RPD free → $35/1k | Shared across Flash tiers |
| Google Maps Tool | 1,500 RPD free → $25/1k | 10,000 RPD free |
| Code Execution | Free | Within supported models |
| URL Context | Standard token pricing | — |
Mid-tier model comparison as of April 2026. Prices per 1M tokens, USD.
| Provider | Model | Input | Output | Strengths |
|---|---|---|---|---|
| Gemini 2.5 Flash | $0.30 | $2.50 | Speed, multimodal, Google ecosystem | |
| Gemini 2.5 Pro | $1.25 | $10.00 | Complex reasoning, long context | |
| OpenAI | GPT-4o | ~$2.50 | ~$10.00 | General capability, broad tooling |
| Anthropic | Claude Sonnet 4.6 | ~$3.00 | ~$15.00 | Reasoning, coding, long documents |
Key takeaway: Gemini Flash-tier models are among the most cost-efficient options for high-volume, latency-sensitive workloads. For complex reasoning tasks, Gemini 2.5 Pro and Gemini 3 Pro are price-competitive with GPT-4o and Claude Sonnet. Model routing — sending simple queries to Flash-Lite and escalating only when needed — is where most cost savings are captured in practice.
Flash-Lite costs 12.5× less per input token than 2.5 Pro. For summarization, classification, extraction, and simple Q&A, it performs comparably. Benchmark your specific tasks, then set routing rules. Revisit quarterly as model capabilities improve.
Batch processing delivers a flat 50% discount on every request with up to 24-hour latency. If a job doesn't need a real-time response — ingestion pipelines, bulk report generation, async enrichment — batch mode is the single highest-leverage cost lever.
Workflows that repeatedly pass the same large documents (policy guides, product catalogs, codebases) should use explicit context caching. Pay once, reuse across many calls at discounted rates. Monitor storage hours to validate the tradeoff.
Long-running agentic workflows accumulate context fast. Every turn that carries over irrelevant history pushes closer to the 200k token pricing cliff — where 2.5 Pro input costs double. Automate context pruning to include only the minimum needed per call.
Implement a two-tier inference pattern: classify incoming requests with a lightweight model, then escalate to Pro only when the classification signals complex reasoning is needed. This is especially effective in chatbots, customer support, and content pipelines with mixed query types.
Search grounding charges ($14–$35 per 1,000 calls beyond the free tier) can compound quickly in high-volume production apps. Set spend alerts specifically for grounding usage, separate from token costs.
Retry storms from poor error handling can double the effective API spend without producing more successful requests. Implement exponential backoff, validate inputs client-side before submission, and alert on elevated error rates.
The right mix of Pro/Ultra subscriptions versus pay-as-you-go API access shifts as usage patterns mature. Teams with predictable, moderate per-user activity often save with subscriptions; high-volume or bursty workloads favor metered API access. Model both scenarios against actual usage before renewing.
Is Google Gemini free? Yes. The Gemini app is free with access to Gemini 2.5 Flash, 100 monthly AI credits, and 15 GB of storage. Free-tier access to the API is also available with rate limits via Google AI Studio.
How much does Gemini 2.5 Pro cost? Gemini 2.5 Pro costs $1.25 per 1M input tokens (up to 200k context) and $10.00 per 1M output tokens. For prompts exceeding 200k tokens, input rises to $2.50/1M and output to $15.00/1M.
How much does Gemini 3 Pro cost? Gemini 3 Pro costs $2.00 per 1M input tokens (≤200k context) and $12.00 per 1M output tokens. Above 200k context, input rises to $4.00/1M and output to $18.00/1M.
What is the cheapest Gemini API model? Gemini 2.5 Flash-Lite is the lowest-cost option at $0.10/1M input tokens and $0.40/1M output tokens. With batch mode, those rates drop to $0.05 input / $0.20 output.
Does Gemini charge for thinking/reasoning tokens? Yes. Gemini charges for both visible output tokens and internal reasoning ("thinking") tokens. Complex prompts that trigger extended chain-of-thought reasoning can significantly increase output costs even when the final visible response is short.
What is the difference between Gemini Pro and Gemini Flash? Pro models prioritize reasoning quality and are suited for complex, multi-step, or multimodal tasks. Flash models prioritize speed and cost efficiency for high-volume, latency-sensitive workloads. Flash-Lite goes further, optimizing for maximum throughput at minimum cost.
How does Gemini batch mode work? Batch mode processes requests asynchronously with no SLA and up to 24-hour latency, in exchange for a 50% discount on all billable tokens. It's available across most Gemini 2.5 and 3 models and is ideal for any non-real-time workload.
How does Gemini pricing compare to OpenAI? Gemini 2.5 Flash ($0.30 input / $2.50 output) is significantly cheaper than GPT-4o (~$2.50 input / ~$10.00 output) for most workloads. At the high end, Gemini 3 Pro and GPT-4o are broadly comparable in price per token for complex tasks.
How do I track Gemini API spend across teams? Google Cloud and Vertex AI provide billing dashboards, but cost allocation across teams, products, and environments typically requires additional tooling. Platforms like Finout integrate Gemini spend into a unified FinOps view alongside cloud, Kubernetes, and other AI providers.
As Gemini usage spreads across engineering, product, and data teams, API costs become harder to attribute and govern. Billing appears as a single line in Google Cloud invoices. Teams can't see which products, features, or environments are driving spend. FinOps and finance can't allocate costs without tagging — and tagging AI workloads rarely happens without automation.
Finout is built for exactly this challenge. It's the FinOps platform for enterprises that have outgrown spreadsheets and fragmented tooling — giving engineering and finance a shared source of truth for cloud, AI, and Kubernetes costs.
For Gemini specifically, Finout provides:
AI-powered cost allocation with Virtual Tags — Automatically allocate Gemini API spend to teams, products, or environments based on usage patterns, without waiting on manual tagging or engineering pipelines. AI Virtual Tags generate allocation rules you can approve in one click and update as fast as your org changes.
Unified visibility via MegaBill — See Gemini costs alongside AWS, GCP, Azure, Kubernetes, Snowflake, Datadog, and other providers in one dashboard. Understand total AI spend — not just Gemini in isolation.
Unit economics — Track cost per inference, cost per feature, or cost per customer. Connect Gemini spend to business outcomes so teams can make decisions based on value, not just raw usage.
Anomaly detection — Get alerted when Gemini costs spike unexpectedly, before they compound into a large monthly surprise.
Shared cost handling — Accurately split AI infrastructure costs shared across teams or workstreams, with configurable allocation logic that finance and engineering both trust.
In the agentic era, AI workloads shift weekly. Prompt architectures change, model versions get upgraded, and new services get integrated. The teams that control costs don't just monitor spend — they build cost systems that adapt as fast as the infrastructure does.
This article is part of Finout's series on AI provider pricing.
Related guides: