Short answer: Google’s Gemini image family now spans three price and quality tiers. At standard API rates, a 1K image costs about $0.034 with Nano Banana 2 Lite, $0.067 with Nano Banana 2, and $0.134 with Nano Banana Pro. Lite is the cheapest option; Pro costs more for stronger control and complex production work.
Prices checked against Google’s developer pricing page on 28 July 2026. API prices can change, so verify the official page before budgeting a large workload.
Gemini image pricing at a glance
| Model | API model ID | Approx. standard cost per 1K image | Best fit |
|---|---|---|---|
| Nano Banana 2 Lite | gemini-3.1-flash-lite-image |
$0.034 | Rapid drafts, high-volume pipelines and cost-sensitive iteration |
| Nano Banana 2 | gemini-3.1-flash-image |
$0.067 | Balanced quality, speed and general production work |
| Nano Banana Pro | gemini-3-pro-image |
$0.134 for 1K/2K | Complex prompts, stronger control and high-value final assets |
The per-image figures are more useful for budgeting than text-output token prices alone. Google separately charges for input, text/thinking output and generated-image tokens. Resolution also changes the final cost: for example, Google lists higher prices for 2K and 4K output from Nano Banana 2.
What changed in June 2026?
On 30 June, Google introduced Nano Banana 2 Lite, its fastest and most cost-efficient Gemini image model. Google positions it as the replacement for the original Nano Banana model in workflows where speed, scale and cost matter most.
Nano Banana 2 and Nano Banana Pro were already the balanced and premium options. The useful change is therefore not simply that “Gemini got cheaper”; developers now have a clearer three-tier routing decision for drafts, everyday production and high-control work.
Which Gemini image model should you use?
Choose Nano Banana 2 Lite for volume
Lite is the sensible first choice for thumbnails, ideation, batch variations and interactive tools where users expect a quick response. Google reports roughly four-second text-to-image output and a standard price of about $0.034 for a 1K image. Use it when the cost of another iteration matters more than getting the strongest result on the first attempt.
Choose Nano Banana 2 for balance
Nano Banana 2 is the generalist model. It offers stronger quality and world knowledge than the Lite tier while remaining cheaper than Pro. It is the safer default for teams that want one model for most image-generation and editing requests without building a complicated router.
Choose Nano Banana Pro for difficult final work
Pro is designed for complex professional use cases where control, reasoning and accuracy are more important than latency. At about $0.134 for a standard 1K/2K image, it is roughly twice the listed 1K cost of Nano Banana 2 and about four times the Lite price. That premium can be justified for a final client asset, but not necessarily for early drafts.
How to estimate your real monthly cost
- Estimate images generated per month, including rejected attempts.
- Separate draft, everyday and final-production requests.
- Multiply each group by the relevant per-image rate and resolution.
- Add text/thinking output, input tokens, search grounding and storage where your workflow uses them.
- Run a small paid test before forecasting at scale because prompt complexity affects iteration count.
For example, 10,000 standard 1K images would be about $340 on Lite, $670 on Nano Banana 2 or $1,340 on Pro before other token and service charges. That simple comparison also shows why routing drafts to Lite and reserving Pro for selected outputs can matter more than switching every request to one model.
Limits of a price-only comparison
The cheapest generated image is not always the cheapest usable image. A model that needs four attempts can cost more in generation fees and staff time than a model that produces the required result in one or two attempts. Quality, latency, editability, subject consistency, text rendering, policy limits and commercial terms should all be tested against your real prompts.
Use our AI tool evaluation checklist to score those trade-offs. For the other model changes published in the same week, see our explainers on Claude Sonnet 5 pricing and plan access and the US export controls affecting Claude Fable 5 and Mythos 5.
How much does Gemini 3 Image cost?
What’s the difference between Flash Image and Pro Image?
Can I use Gemini image models for commercial content?
Is Gemini cheaper than other AI image generators?
Primary sources
- Google AI for Developers: Gemini API pricing — current token and per-image rates.
- Google: Nano Banana 2 Lite launch — positioning, speed and model-family roles.
- Google: building with Nano Banana 2 — capabilities and developer access.
- Gemini API release notes — model availability and lifecycle changes.
Update log: 28 July 2026 — corrected the headline comparison to include Nano Banana 2 Lite, replaced incomplete text-token pricing with per-image costs, added a three-model decision table, budgeting method, limitations and primary-source links.