MiniMax H3 Image Highspeed Edit: 16 Refs, $0.0219

The short answer
MiniMax H3 Image Highspeed Edit, the editing endpoint of the Highspeed variant of MiniMax H3 Image, is now callable through a model API. It combines up to 16 reference images with text instructions and outputs up to 20 megapixels. It runs in the AdGraam Playground from $0.0219 per image at 1024×1024.
On this page (10 sections)
What is MiniMax H3 Image Highspeed Edit?
MiniMax H3 Image Highspeed Edit is the image editing endpoint of the Highspeed variant of MiniMax H3 Image. It transforms and combines up to 16 reference images using text instructions, with custom output sizes up to 20 megapixels. It is callable on a model API and runs in the AdGraam Playground.
MiniMax H3 Image is the image member of the company's H3 family, separate from the H3 video model. A third-party write-up from Omniakey, citing the Hugging Face model card, says one 33.1B-parameter checkpoint handles both text-to-image and reference editing, with weights on Hugging Face under a custom community license. That detail comes from the write-up, not from a MiniMax page we can quote directly.
For earlier coverage of the sibling endpoints, see our MiniMax H3 Image Edit and Highspeed text to image articles.
MiniMax H3 Image Highspeed Edit release date and availability
The edit endpoint was first seen on a public model API on 2026-11-11. The Highspeed and standard H3 Image models appear to have launched a day or so earlier: the Creative AI Show uploaded a comparison on 2026-11-10 saying both had been released.
Availability today:
- AdGraam Playground, for both Edit and Text to Image
- A model API, per the registry listing
- Per Omniakey, MiniMax's own API, MiniMax Code and Hailuo AI, plus open weights on Hugging Face
MiniMax H3 Image Highspeed Edit key features and specs
The edit model takes 1 to 16 reference images and a text prompt, returns 1 to 10 images per request, and preserves the first image's aspect ratio by default. Outputs reach 20 megapixels with custom sizes, in jpeg, png, webp or jpg.
Designer desk with stacked reference photos and color swatches
| Spec | Details |
|---|---|
| Model | MiniMax H3 Image Highspeed Edit |
| Maker | MiniMax |
| Reference images | Up to 16 |
| Max output | 20 megapixels (custom sizes) |
| Image size presets | Square HD, Square, Portrait 3:4, Portrait 9:16, Landscape 4:3, Landscape 16:9, Auto |
| Images per request | 1 to 10 |
| Output formats | jpeg, png, webp, jpg |
| AdGraam price, 1024×1024 | About $0.0219 per image |
| AdGraam price, 2048×2048 | About $0.053 per image |
Omniakey reports that the standard and Highspeed IDs share parameters and limits, and that masked edits and transparent backgrounds are not supported. If you need an alpha cutout, use a different model. Omniakey also reports that MiniMax claims 4K generation in under 10 seconds, without naming a baseline. Treat that as a vendor claim.
MiniMax H3 Image Highspeed Edit pricing and free access
In the AdGraam Playground, Highspeed Edit costs about $0.0219 per image at 1024×1024 and about $0.053 per image at 2048×2048. The Text to Image variant has the same rates. See pricing for credit packs.
Other platforms differ. Runware lists a reference-to-image example at $0.074259 per run at 2K, and a text-to-image magazine cover example at $0.04219. The Creative AI Show, paying through a different API, measured roughly 3.8 to 4.1 cents per Highspeed image against about 2.2 to 2.3 cents for standard H3 Image. No free tier is announced.
MiniMax H3 Image Highspeed Edit vs standard H3 Image
Highspeed is not automatically the better pick. In the Creative AI Show comparison, the standard model followed prompts more closely and cost less on their provider, while Highspeed finished in about 10 to 12 seconds against roughly 12 to 14.
Their results were mixed. Highspeed did best on one physics-heavy prompt where standard failed, and failed on a pastry chef scene. They also judged Nano Banana 2.1 more realistic than H3 on that scene. This is one informal test by two creators with a handful of prompts, so run your own product prompts before committing.
| Model | Cost per image (Creative AI Show) | Time per image | Prompt adherence (their view) |
|---|---|---|---|
| H3 Image | About 2.2 to 2.3 cents | About 12 to 14 sec | Stronger overall |
| H3 Image Highspeed | About 3.8 to 4.1 cents | About 10 to 12 sec | Mixed |
How to use MiniMax H3 Image Highspeed Edit for ad creatives
Sixteen reference slots give you room to keep a campaign looking consistent. A product shot, a brand style frame, a model reference and a background can all go into one request. The prompt refers to each by position.
Product bottles shown in several styled ad scenes
Practical uses:
- Place a product from one image into lifestyle scenes from others, then output 9:16 for Reels and Stories.
- Generate up to 10 variants per request for angle or background testing.
- Resize from one source to Square, 3:4 and 16:9 for placement coverage.
Chain it with upscaling or video in the Flow Builder, or build statics from a product URL in the Creative Generator. Our static ads guide covers the fundamentals. Compare it with FLUX 3 Image Edit, covered in our FLUX 3 article, in the Playground.
What to watch next
MiniMax has not published edit-specific benchmarks in the material available, and the early speed numbers come from launch week, when load can skew timing. The Creative AI Show says it plans to test edit mode in a future video.
Watch for independent tests of text rendering and product fidelity across multiple references, plus clarity on commercial terms for outputs.
Key takeaways
- MiniMax H3 Image Highspeed Edit accepts up to 16 reference images plus a text instruction.
- Custom output sizes go up to 20 megapixels, and the first image's aspect ratio is preserved by default.
- In the AdGraam Playground it costs about $0.0219 per image at 1024×1024 and $0.053 at 2048×2048.
- One request can return 1 to 10 images in jpeg, png, webp or jpg.
- A third-party write-up from Omniakey, citing the model card, describes H3-Image as a 33.1B-parameter model with open weights under a custom community license.
- An early creator comparison found the standard model followed prompts better than Highspeed, and Highspeed was not much faster.
Compare 200+ AI video and image models side by side
Run Kling, Veo, Sora, Seedance, Flux, Ideogram and Recraft in one Playground, or turn any product URL into ready-to-launch ads.
Frequently asked
When did MiniMax H3 Image Highspeed Edit come out?
The edit endpoint was first seen as callable on a model API on 2026-11-11. A YouTube comparison from the Creative AI Show was uploaded on 2026-11-10 and says H3 Image and H3 Image Highspeed had just been released.
How much does MiniMax H3 Image Highspeed Edit cost?
In the AdGraam Playground it is about $0.0219 per image at 1024×1024 and about $0.053 per image at 2048×2048. Pricing on other platforms varies.
How many reference images can it use?
Up to 16. You describe in the text prompt how they should be combined or changed, and the first image's aspect ratio is kept by default.
Is Highspeed better than standard H3 Image?
Not clearly. The Creative AI Show found the standard model followed prompts better and cost less on one provider, while Highspeed was only about 1 to 3 seconds faster in their runs. That is a small, informal comparison.
Can I use it for commercial ads?
Terms for generated outputs are not stated in the available material. MiniMax's open weights use a custom community license, so check it before commercial use of the weights.
Sources
- MiniMax H3 IMAGE vs Nano Banana - We Compared High ...youtube.com
- MiniMax H3 Image Highspeedrunware.ai
- MiniMax H3-Image: 4K Images, Open Weights and API Accessomniakey.com
More AI model news & creative guides

MiniMax H3 Image Highspeed: 20MP, $0.0219/Image
MiniMax H3 Image Highspeed text to image is live: sizes up to 20MP, from $0.0219 per image. Get the spec table, ad workflow breakdown and access details.

Midjourney V8.2 Style Transfer: 4 Outputs, $0.129
Midjourney V8.2 Style Transfer restyles an image and returns 4 results for $0.129 per run in AdGraam Playground. Get the spec table, ad workflow and access.

Bria Fibo 2 Text to Image: Specs & $0.04 Price
Bria Fibo 2 (Fibo 2.0) text to image runs at $0.04 per image in AdGraam Playground. Get the spec table, ad workflow breakdown and access details.
Best AI Ad Generators for D2C Brands in 2026
Ten AI ad generators compared for D2C and e-commerce brands: what each makes, real starting prices, free options, and which suits Indian brands paying in rupees.
Break Even ROAS for COD Orders: How RTO Changes the Math
How to work out break even ROAS when a share of COD orders comes back as RTO, with worked rupee examples, a prepaid and COD mix, and why Ads Manager ROAS overstates COD.
What Indian D2C Brands Run on Meta: 2,790 Ads Analyzed
We studied 2,790 Facebook and Instagram ads from 14 Indian brands and their 747 longest-running ads: formats, angles, offers, languages and CTAs that last.
Your move
See your product as a winning ad
Turn a product URL into static ads and UGC-style videos, rebuild the competitor ads that already work, and read your Meta data to know what to make next.
AdGraam News DeskAI video and image news
The AdGraam News Desk (part of AdGraam, also searched as Adgram or Adgram AI) tracks AI image and video model launches, leaks and rumors from labs, leakers and creators, and checks each claim against its sources. Unconfirmed stories say so, and each page is updated as its story develops.