- Blog
- MiniMax H3 Image Pricing: Official API Rates & Cost Guide
MiniMax H3 Image Pricing: Official API Rates & Cost Guide
on 8 days ago

MiniMax H3 Image pricing uses separate token charges for text input, reference images, and generated images. The Standard model has lower image token rates than Highspeed, while both charge the same rate for text input. Your bill depends on usage, so a quoted example price per image is a planning reference, not a fixed tariff.
This guide explains the official rates, a temporary 4K offer, and practical budgets for developers and creative teams. Prices were checked on November 12, 2026. All amounts are USD; calculations describe direct MiniMax API usage, not Shotyard credits. The illustrations are conceptual, not samples used to evaluate the model.
Official MiniMax H3 Image pricing: Standard versus Highspeed
The image section of the official MiniMax pricing overview lists these rates per million tokens:
| Model | Text input | Image input | Image output |
|---|---|---|---|
| MiniMax-H3-Image | $2.30 | $1.50 | $3.80 |
| MiniMax-H3-Image-Highspeed | $2.30 | $3.00 | $7.60 |
Use the complete model name when comparing quotes. A price for an H3 video task or an older image model does not establish the price of H3 Image.
For identical token counts, H3 Image Highspeed doubles the image components, not the entire bill. The text charge stays unchanged. Whether its speed is worth paying for requires a test of your own workload; this guide makes no latency claim.
Calculate image generation cost from token usage

Calculate each billed category separately, then add the results:
Cost = (text tokens × text rate + image input tokens × image input rate + image output tokens × image output rate) ÷ 1,000,000.
For a hypothetical Standard request with 1,000 text tokens, no reference image, and 6,000 output tokens, the arithmetic is:
- Text: 1,000 × $2.30 ÷ 1,000,000 = $0.00230.
- Output: 6,000 × $3.80 ÷ 1,000,000 = $0.02280.
- Total: $0.02510 before any eligible promotion.
At Highspeed rates, the same assumed counts produce $0.00230 + $0.04560 = $0.04790. Adding 2,000 reference-image tokens would add $0.00300 on Standard or $0.00600 on Highspeed.
These token counts are invented solely to demonstrate the formula. They are not measurements, official examples, or predictions for a particular resolution. Replace them with actual usage before estimating production spend. Keep the full decimal values during multiplication and round only the final budget; rounding every small request upward can noticeably inflate a large forecast.
Official example costs and the temporary 4K discount
MiniMax publishes illustrative text-to-image totals of $0.02179 for Standard at 1024 × 1024 and $0.03529 for Highspeed at that size. The examples can vary with text usage, including prompt enhancement. They are not guaranteed quotes for every request at those dimensions.
The pricing overview's 4K offer gives both models 50% off total request cost when the delivered image has at least 16,600,000 pixels, ending November 17, 2026, at 16:00 UTC, exclusive. Treat the offer as temporary and check it again before purchasing.
Pixel count matters more than a casual “4K” label. Multiplying 3840 × 2160 gives 8,294,400 pixels, below that threshold; 4096 × 4096 gives 16,777,216, above it. Both calculations describe dimensions, not a promise that any particular request qualifies.
Build your recurring budget at the regular rates. If a qualifying request would otherwise cost $0.08, a 50% discount would reduce it to $0.04. That is a hypothetical illustration of the discount, not another published MiniMax price.
What changes image editing cost?
References, output settings, and prompt handling belong in the same estimate. The official image generation guide documents 1–16 reference images for edits and 1–10 output images per request. Text-to-image uses prompt enhancement by default; edits always use it. Inspect the revised prompt when investigating unexpected usage.
For dimensions, the guide offers explicit width and height or an aspect ratio plus an area level. Its 2K and 16:9 example produces 2720 × 1536 pixels. Record the returned dimensions rather than assuming that a short size label identifies the longest edge.
In a pilot, change one variable at a time. Compare one reference against three only if both versions can satisfy the same brief. A reference that prevents several unwanted generations may be useful, even though reference inputs add a billable category. Conversely, a decorative reference that contributes nothing to the deliverable deserves removal.
Record why an image passed review alongside its cost. A note such as “correct bottle silhouette after two attempts” gives the next person preparing a budget something concrete to work with.
Budget for accepted images, including revisions

For a campaign budget, count the images your team can actually use. Calculate spend divided by accepted images alongside the cost of each generation.
Suppose a project needs 200 approved images and you budget for three attempts each. That means 600 generations. Using the hypothetical token totals above, the generation allowance becomes:
| Planning scenario | Attempts | Assumed cost each | Total |
|---|---|---|---|
| Standard, no reference input | 600 | $0.02510 | $15.06 |
| Highspeed, no reference input | 600 | $0.04790 | $28.74 |
If each approved image also requires one Standard edit with the assumed 2,000 input-image tokens, each edit would cost $0.02810. Another 200 edits add $5.62, bringing that Standard scenario to $20.68. This example assumes every request has the stated token usage and applies no promotion. Dividing $20.68 by 200 approved images gives $0.1034 per accepted asset before the reserve.
A 20% planning reserve would make the allowance $24.816, or $24.82 rounded. That reserve is your budgeting choice, not a MiniMax fee. Keep staff review time, storage, delivery, and any separately applicable charges outside this inference subtotal.
After the pilot, replace “three attempts” with observed results. Track simple background images separately from demanding product compositions, and agree on acceptance criteria before judging the results. Otherwise, a low average cost can hide work that never reaches approval.
Compare API charges with a Shotyard workflow
MiniMax API pricing and an application subscription are different purchasing decisions. Review Shotyard pricing for the application's current offer; the dollar figures above do not establish its credit conversion or model availability.
If you are comparing image providers, the GPT Image 2.6 pricing guide provides another budgeting framework. Compare the same finished deliverable, including revisions, rather than treating a low headline rate as the whole decision.
For occasional work, include the time needed to set up an integration. For recurring production, keep a request log with the model, settings, billed usage, outcome, and acceptance decision. Reconcile a small batch against account billing before scaling it. Use those reconciled costs in the next forecast.
FAQ
Is H3 Image Highspeed cheaper than Standard?
Its published image token rates are higher. With identical usage, the calculation above makes Standard cheaper. A production decision should also consider whether a tested workflow saves enough time or revisions to justify the difference.
Does every image described as 4K receive the discount?
No. Check the delivered pixel area and the offer deadline. The 3840 × 2160 calculation above falls below the stated threshold. After the promotion ends, use regular rates unless MiniMax announces a replacement.
Can I assume free generations or free retries?
Do not include either in a budget without explicit applicable terms. The official guide's SDK example disables automatic retries to avoid duplicate generation and charges. An uncertain response should prompt investigation before you submit the same paid job again.
Conclusion
To budget MiniMax H3 Image pricing, start with the published token rates and record what each approved asset takes to produce. Run a small representative batch, include revisions, and forecast at regular rates. Before applying the temporary discount, check the delivered dimensions and the deadline.
