MiniMax H3 Image
ImageMiniMax · Text to image · from $0.03
MiniMax H3 Image Text-to-Image generates high-quality images from text prompts with flexible aspect ratios and built-in prompt enhancement for creative visuals, product imagery, marketing assets, and production workflows. Ready-to-use REST inference API, best performance, no coldstarts, affordable pricing.
MiniMax H3-Image is an image model released 2026-11-10 (distinct from the H3 video model): native 16 MP (5152×3872), 4K-class output in 10–20 seconds, up to 16 reference images, open weights.
- Native 16 MP output, 5152×3872 per the official figures — no small-image-then-upscale step.
- Officially 10–20 seconds for a 4K×4K-class image.
- Up to 16 reference images, with multi-reference and editing support.
- Open weights and self-hostable; also available via API, MiniMax Code and Hailuo AI.
- From $0.03 per run upstream; the launch-week 50% discount on 16MP+ runs ends 2026-11-18.
Worth knowing
Name collision: MiniMax H3 is the video model open-sourced 2026-07-31; H3-Image is the image model released 2026-11-10; P-H3-Image is Pruna’s distilled build (8 steps, 2–16 MP, 8 references), not the official model.
Testers report strong portrait realism but contested aesthetics; public data on stability beyond 4 reference images is thin.
This model is not yet available in the Draftio canvas. This page summarizes upstream capabilities, official samples and billing; similar models are available in the canvas today.

PromptExpand
A photorealistic close-up portrait of an elderly fisherman with weathered skin, a grey beard and a dark wool cap, standing on a harbor dock on an overcast morning, soft natural light, shallow depth of field
Parameters
Official parameters only. Explicit capabilities are limited to what the official description states.
| Dimension | MiniMax H3 Image |
|---|---|
| Task | Text to image |
| Starting price | $0.03/run |
| Resolution | 1k / 2k / 4k |
| Aspect ratio | 1:1 / 1:2 / 2:1 / 1:3 / 3:1 / 2:3 +9 |
| Duration | — |
| Explicit capabilities | Text to image |
Explicit capabilities are only those stated in the official description. Nothing is inferred.
Pricing
Beyond the starting price, resolution and duration change the actual cost.
| Model | Unit price | Resolution / duration | Source | Verified |
|---|---|---|---|---|
| MiniMax H3 Image | $0.03/run | 1k | wavespeed | 11/11/2026 |
Prices are official starting prices, excluding tax and FX. Refer to the official source.
Samples for this task
Official samples. Task and parameters are defined by the original page.

A photorealistic close-up portrait of an elderly fisherman with weathered skin, a grey beard and a dark wool cap, standing on a harbor dock on an overcast morning, soft natural light, shallow depth of field

A photorealistic flat-lay breakfast on a light wooden table: two golden croissants, a glass of orange juice, a white ceramic coffee cup, a linen napkin and a small vase with wildflowers, soft window light from the left

A vintage-style travel poster of a white lighthouse on a rocky coast at sunset, with the English title SEA BREEZE in bold cream letters at the top and no other text

A photorealistic wide landscape of a misty mountain lake at sunrise, a small wooden rowing boat near the shore, pine forest, soft golden light on the water

A photorealistic Scandinavian bedroom in soft morning light, a bed with white linen, a wooden bedside table with a small lamp, a large window with sheer curtains, a green plant in the corner

A glamorous young female singer standing inside a retro-futuristic recording studio, wearing an elegant glossy black dress whose layered circular skirt resembles the grooves of a vinyl record. She holds a chrome microphone while large transparent record discs hang behind her like sculptural art. Strong full-body silhouette, sophisticated black, silver, and deep red palette, cinematic stage lighting, realistic vinyl, metal, skin, and fabric, high-fashion music editorial.
Other models of the same type
Same type, ready for a side-by-side comparison. This is not a ranking.
Data source
Prices and samples come from the wavespeed official model page; last verified 11/11/2026. This page currently includes 6 official samples. Prompts and parameters are defined by the original page.
Prices change with resolution, duration and official updates. This page shows starting prices and default configurations only; refer to the official source for actual cost.
FAQ
- What is the difference between MiniMax H3, H3-Image and P-H3-Image?
- H3 is the video model (open-sourced 2026-07-31, 2K/15s, native stereo). H3-Image is the image model released 2026-11-10 (16 MP, 16 references). P-H3-Image is Pruna’s distilled build of H3-Image: faster and smaller, with 8 references.
- How many reference images are stable?
- The official cap is 16. Testers report more subject/style bleed as references increase, with no published threshold; start with 2–4 and add gradually for your assets.
- H3-Image vs Seedream 5.5 — which should I use?
- Use H3-Image for native 4K, very large pixel counts or many references. Use Seedream 5.5 for multi-round edits that must not drift, transparent assets, or a lower per-image cost.
- How much does MiniMax H3 Image cost?
- From $0.03 per run (tiering across 1K/2K/4K follows the official price sheet); 16MP+ runs are 50% off during launch week, ending 2026-11-18. Last verified: 11/11/2026.
- Does MiniMax H3 Image support Chinese text?
- The official description does not explicitly list Chinese text capability. Validate with your actual copy before using it for Chinese text.
- What is the official task category for MiniMax H3 Image?
- The official task category is Text to image (Text to Image API). This page only shows verifiable tasks, prices and parameters; it does not infer use cases.
- How should I choose between MiniMax H3 Image and similar models?
- Start with a side-by-side comparison of MiniMax H3 Image and Z Image Turbo: price, parameters and official samples on one page.


