AI Image API Cost Calculator
Image APIs bill per picture — but the picture you ship is rarely the first one generated. Set your monthly volume and how many attempts one usable image takes, and compare what each model really costs you.
gemini-3.1-flash-image (Nano Banana 2) shuts down on October 29, 2026 — 17 days from today. The replacement, Nano Banana 2.1, is half the price at every shared resolution. Migration guide & countdown → All models compared at your settings
| Model — plan | Per image | Per usable image | Monthly | Yearly |
|---|
Sizes differ between vendors (1K vs 1024×1024 vs ~1MP) — close enough for budgeting, not for pixel-counting. Each model uses the closest available size to your selection and its default plan.
Full published per-image rate table
Output prices per generated image, verified 2026-10-12. Token-billed models also charge input tokens — for edit-heavy workloads with many reference images, that line is not optional.
Nano Banana 2.1 · Google · API id: gemini-nano-banana-2.1
Google's current workhorse image model (GA Oct 6, 2026) and the designated replacement for gemini-3.1-flash-image. Output billed per token: a 1K image is 1,120 output tokens at $30 per million. Input tokens tripled vs the predecessor ($1.50/1M), which matters for edit-heavy flows with many reference images.
| Plan | 1K | 2K | 4K |
|---|---|---|---|
| Standard | $0.034 | $0.050 | $0.076 |
| Batch API (−50%) | $0.017 | $0.025 | $0.038 |
Source: Gemini API pricing page (Google), October 2026. Rate source ↗ · Open the Nano Banana 2.1 calculator →
Nano Banana 2 (retiring Oct 29) · Google · API id: gemini-3.1-flash-image
DEPRECATED — shuts down October 29, 2026. Migrate to Nano Banana 2.1 (gemini-nano-banana-2.1): output prices halve at every shared resolution, but the 0.5K tier disappears and input tokens cost 3× more. Listed here so existing users can price the migration.
| Plan | 0.5K | 1K | 2K | 4K |
|---|---|---|---|---|
| Standard | $0.045 | $0.067 | $0.101 | $0.151 |
| Batch API (−50%) | — | $0.034 | $0.051 | $0.075 |
Source: Gemini API pricing page (Google), October 2026. Rate source ↗ · Open the Nano Banana 2 (retiring Oct 29) calculator →
Nano Banana Pro · Google · API id: gemini-3-pro-image
Google's premium image tier for complex visual tasks. 1K and 2K cost the same; 4K jumps to $0.24.
| Plan | 1K | 2K | 4K |
|---|---|---|---|
| Standard | $0.134 | $0.134 | $0.240 |
| Batch API (−50%) | $0.067 | $0.067 | $0.120 |
Source: Gemini API pricing page (Google), October 2026. Rate source ↗ · Open the Nano Banana Pro calculator →
Nano Banana 2 Lite · Google · API id: gemini-3.1-flash-lite-image
The lite tier of the Gemini image line, positioned for fast stills before animating them. Single headline rate at 1K; treat as approximate and confirm in the pricing page for other sizes.
| Plan | 1K |
|---|---|
| Standard (≈1K) | $0.034 |
Source: Google launch coverage, June 30, 2026. Rate source ↗ · Open the Nano Banana 2 Lite calculator →
GPT Image 2.5 · OpenAI · API id: gpt-image-2.5
OpenAI's current image models (Flare and Sunburst share token rates). Billed per token; figures below are the image-output cost for a 1024×1024 image at each quality tier — prompt and reference-image input tokens are extra. Quality moves price far more than size does.
| Plan | 1024×1024 |
|---|---|
| Low quality · 1024×1024 | $0.0060 |
| Medium quality · 1024×1024 | $0.013 |
| High quality · 1024×1024 | $0.053 |
| X-High quality · 1024×1024 | $0.094 |
| Max quality · 1024×1024 | $0.211 |
Source: OpenAI API pricing page, per-image output costs as published Sep–Oct 2026. Rate source ↗ · Open the GPT Image 2.5 calculator →
FLUX.1 family · Black Forest Labs · API id: flux-1.1-pro / flux-1-dev / flux-1-schnell
Flat per-image pricing on the BFL API (roughly 1MP outputs). Schnell is the volume workhorse; note FLUX.1 [dev] carries a non-commercial license — the per-image price is not the whole cost picture.
| Plan | ~1MP |
|---|---|
| FLUX.1 [schnell] | $0.0030 |
| FLUX.1 [dev] (non-commercial license) | $0.025 |
| FLUX 1.1 [pro] | $0.040 |
Source: Black Forest Labs API pricing. Rate source ↗ · Open the FLUX.1 family calculator →
Grok Imagine Image · xAI · API id: grok-imagine-image
Flat $0.02 per generated image on the xAI API; edits and composites add $0.002 per input image.
| Plan | flat |
|---|---|
| xAI API (direct) | $0.020 |
Source: xAI API pricing, as documented Oct 2026. Rate source ↗ · Open the Grok Imagine Image calculator →
How to read these prices
Batch is the cheapest lever you have. Google's Batch API halves image output prices across the whole Nano Banana line. If your images are not needed interactively, nothing else you optimize will beat a flat 50%.
Quality tiers move price more than size. On GPT Image 2.5, going from medium to max quality at the same 1024×1024 multiplies the output cost roughly 16×. Draft at low, finish at high — and set size and quality explicitly, because auto has no predictable price.
Count attempts, not images. A $0.034 model you re-roll three times costs more than a $0.053 model you accept on the first try. That is why this calculator prices the usable image, not the generation.