# Pricing reference (April 2026)

Per-image cost matters because iteration discipline depends on it. Numbers below are list price; batch APIs and tier discounts may apply separately.

## Per-image cost — head-to-head

| Model | 0.5K | 1K / 1024² | 2K | 4K |
|---|---|---|---|---|
| Gemini Flash 3.1 | $0.045 | $0.067 | $0.101 | $0.151 |
| Gemini Pro 3 | n/a | $0.134 | $0.134 | $0.24 |
| OpenAI gpt-image-1.5 (low) | n/a | $0.009 | n/a | n/a |
| OpenAI gpt-image-1.5 (medium) | n/a | $0.034 | n/a | n/a |
| OpenAI gpt-image-1.5 (high) | n/a | $0.133 | n/a | n/a |

OpenAI sizes are fixed: 1024×1024, 1024×1536 (portrait, +50%), 1536×1024 (landscape, +50%). No 2K/4K.

## Common scenarios — total cost calculator

### Logo project (exploration → final)

| Step | Calls | Model | Cost each | Subtotal |
|---|---|---|---|---|
| Initial exploration | 8 | Gemini Flash 1K | $0.067 | $0.54 |
| Refinement (3 directions × 3 variants) | 9 | Gemini Flash 1K | $0.067 | $0.60 |
| Final delivery (2K, 3 variants) | 3 | Gemini Pro 2K | $0.134 | $0.40 |
| 4K master | 1 | Gemini Pro 4K | $0.24 | $0.24 |
| **Total** | 21 | | | **$1.78** |

vs all-OpenAI workflow:
| Step | Calls | Model | Cost each | Subtotal |
|---|---|---|---|---|
| Exploration `n=4` | 5 | gpt-image-1.5 low | $0.036 | $0.18 |
| Refinement | 9 | gpt-image-1.5 medium | $0.034 | $0.31 |
| Final | 3 | gpt-image-1.5 high transparent | $0.133 | $0.40 |
| **Total** | 17 | | | **$0.89** |

OpenAI is cheaper for logo projects when you want transparent backgrounds. Gemini Pro wins on final-output quality at 2K-4K.

### UI mockup project (mobile + desktop, 5 screens)

| Step | Calls | Model | Cost each | Subtotal |
|---|---|---|---|---|
| Wireframe sketches | 5 | Gemini Flash 1K | $0.067 | $0.34 |
| Hi-fi 5 mobile screens (9:16, 2K) | 5 | Gemini Pro 2K | $0.134 | $0.67 |
| Hi-fi 1 desktop dashboard (16:9, 4K) | 1 | Gemini Pro 4K | $0.24 | $0.24 |
| 1 device-framed marketing shot | 1 | Gemini Pro 4K | $0.24 | $0.24 |
| **Total** | 12 | | | **$1.49** |

### Marketing hero set (3 hero images for landing page)

| Step | Calls | Model | Cost each | Subtotal |
|---|---|---|---|---|
| Concept exploration | 6 | Gemini Flash 2K | $0.101 | $0.61 |
| Final hero × 3 variants | 3 | Gemini Pro 4K | $0.24 | $0.72 |
| **Total** | 9 | | | **$1.33** |

### Icon set (24 icons, consistent style)

| Step | Calls | Model | Cost each | Subtotal |
|---|---|---|---|---|
| Style locking (3 candidate sets of 6 icons each as grids) | 3 | Gemini Flash 2K | $0.101 | $0.30 |
| Final 24 icons in 4 grids of 6 | 4 | Gemini Pro 2K | $0.134 | $0.54 |
| Re-runs for 3 weak icons | 3 | Gemini Pro 1K (single-icon) | $0.134 | $0.40 |
| Background removal pass via gpt-image-1.5 | 24 | gpt-image-1.5 medium | $0.034 | $0.82 |
| **Total** | 34 | | | **$2.06** |

Or skip the OpenAI background-removal pass and use `rembg` locally for free → $1.24 total.

### Product photography set (10 product shots)

| Step | Calls | Model | Cost each | Subtotal |
|---|---|---|---|---|
| Lighting/angle exploration on 1 product | 5 | Gemini Flash 2K | $0.101 | $0.51 |
| Final 10 product shots, 4K | 10 | Gemini Pro 4K | $0.24 | $2.40 |
| Catalog cutouts (transparent) | 10 | gpt-image-1.5 high transparent 1024² | $0.133 | $1.33 |
| **Total** | 25 | | | **$4.24** |

## Batch API discounts

- **Gemini Batch API:** ≈ 50% off, 24-hour turnaround. Use for non-interactive bulk generation (e.g., generating 100 product variants overnight).
- **OpenAI Batch API:** 50% off, 24-hour turnaround. Same use case.

For interactive design work (the main use of this skill), batch APIs are useless because you need to see results to iterate. They matter for one-shot bulk jobs.

## Cost discipline rules

1. **Default to Flash for first 3 calls.** Only promote to Pro once you've locked the prompt structure.
2. **Don't use Pro for exploration.** $0.13/call × 30 explorations = $4 vs $2 on Flash.
3. **Don't use 4K until the final.** 2K is half the cost and indistinguishable on screen for most use cases.
4. **Don't iterate >3 times on the same image.** Drift compounds, cost compounds. Rewrite the prompt from scratch.
5. **For transparent PNGs, decide upfront:** native via gpt-image-1.5 high ($0.13) or generate-on-white via Gemini + post-process with `rembg` (free, ~5s extra). For one-off finals, native is faster. For batch, post-process is much cheaper.
6. **For 24+ assets in the same style, prefer multi-turn editing** on Gemini Pro over independent calls. Pro's chat-mode keeps style consistent and you only pay full price for the first image; subsequent edits often use less compute.

## Sanity-check budgets per asset type

| Asset | Reasonable budget | Red flag if you exceed |
|---|---|---|
| Single logo, exploration to final | $1-3 | $5 |
| Icon set of 12-24 | $1-3 | $5 |
| 5-screen UI mockup (mobile) | $1-2 | $4 |
| 1-screen hi-fi dashboard | $0.50-1 | $2 |
| 3-image hero set for landing | $1-2 | $3 |
| 10-image product catalog | $3-5 | $8 |

If you're approaching the red flag, pause and ask the user whether to keep going or pivot strategy.