20% offBlack Friday11d 03:20:20 left

H3 Image vs GPT Image 2.6: the same five prompts

We gave MiniMax H3 Image and GPT Image 2.6 the same five prompts and ran each once. Below are the pictures and what each one got right or missed.

Try the prompts yourself

What we saw

Both models got every letter and character right on the three text prompts. They differed on how much of the rest of the brief each picture kept.

  • Chalkboard menu

    MiniMax H3 Image: Matches the brief. Every word right

    GPT Image 2.6: Matches the brief. Every word right, extra flourishes

  • Five-step infographic

    MiniMax H3 Image: Matches the brief. Text right, line icons

    GPT Image 2.6: Partly. Text right, coloured art, not line icons

  • Candid portrait

    MiniMax H3 Image: Partly. Looks at camera, no mole

    GPT Image 2.6: Partly. Hair lighter than asked

  • Film still: starship in the desert

    MiniMax H3 Image: Partly. Hull is a plain cylinder

    GPT Image 2.6: Matches the brief. Reads as a starship

  • Chinese ink landscape with poem and seal

    MiniMax H3 Image: Partly. Text right, faint snow

    GPT Image 2.6: Matches the brief. Text and snow as asked

MiniMax H3 Image: 2 match, 3 partly, 0 missed · GPT Image 2.6: 3 match, 2 partly, 0 missed

Judged against what each prompt asked for. This is our reading of one run each.

  • Where H3 Image stayed closer: the menu has just the two doodles we asked for, and the infographic uses simple line icons as asked.
  • Where GPT Image 2.6 stayed closer: the portrait looks away and has the mole, the starship reads clearly as a starship, and the ink painting has snow-covered mountains. Its portrait hair came out lighter than the dark hair we asked for.
  • Picture size: at the 1K tier GPT Image 2.6 gave a larger picture (1448 × 1086 against 1184 × 880).

Five prompts and one run each is a small test. It shows what these models did here, not which one is better overall.

How to write prompts like these: the H3 Image prompt guide

How we ran it

Each model got each prompt once, with the same words, at its 1K setting and a 4:3 frame, through the KIE API on November 15, 2026, 16:22 UTC. Prompt rewriting was switched off wherever the model has that option. There were no reruns and no retouching. GPT Image 2.6 ran on KIE's economy channel, which may not match every GPT Image 2.6 variant OpenAI sells. Output sizes differ by model: MiniMax H3 Image 1184 × 880, GPT Image 2.6 1448 × 1086, Seedream 5.5 1152 × 864, Qwen Image 2.1 1184 × 896.

Chalkboard menu

Every word in quotes: a title, two columns with prices and a footer line, plus two chalk doodles.

More H3 Image text-in-image prompts
Prompt (same for every model)

A café chalkboard menu hanging on a brick wall, hand-lettered in white and yellow chalk. Title at the top: "THE CORNER CUP". Two columns. Left column under "COFFEE": "Espresso 3.00", "Flat White 4.20", "Oat Latte 4.80", "Cold Brew 4.50". Right column under "BAKERY": "Butter Croissant 3.50", "Banana Bread 3.80", "Cardamom Bun 4.00". At the bottom: "Open daily 7am – 4pm · Free refills on filter coffee". Small chalk doodles of a coffee cup and a leaf. Warm morning light, photo.

MiniMax H3 Image: Chalkboard menuMiniMax H3 Image
1K · Add detail off

Every word matches the prompt, with the cup and leaf doodles asked for.

GPT Image 2.6: Chalkboard menuGPT Image 2.6
1K · economy channel

Every word matches. Adds yellow underlines, steam and ivy that the prompt did not ask for.

Five-step infographic

A title and five equal columns, each with one simple line icon and one caption, and no other text.

More H3 Image prompts for readable text
Prompt (same for every model)

A clean flat infographic poster on a light beige background. Title at the top: "Pour-Over Coffee in 5 Steps". Below it, five equal columns from left to right, each with one simple line icon and one caption: "1. Heat water to 94°C", "2. Grind 20 g coffee", "3. Rinse the filter", "4. Bloom 30 seconds", "5. Pour to 320 g". No other text.

MiniMax H3 Image: Five-step infographicMiniMax H3 Image
1K · Add detail off

Title and all five captions match. The icons are simple lines as asked, but some are hard to read as their step.

GPT Image 2.6: Five-step infographicGPT Image 2.6
1K · economy channel

Title and all five captions match. Coloured illustrations instead of the line icons asked for.

Candid portrait

Hair colour, a mole, a pose, a gaze direction and soft window light, all in one paragraph.

More H3 Image portrait prompts
Prompt (same for every model)

Candid at-home snapshot of a woman in her early thirties with shoulder-length dark hair and a small mole under her left eye, wearing an oversized oatmeal knit sweater, sitting on a windowsill with a mug of coffee, caught mid-laugh as she pushes her hair back with one hand and looks slightly away from the camera. Close-up from the chest up. Soft diffused morning light through a sheer white curtain falls on her face, exposure set for the face, bright high-key image, low contrast, muted colours, fine film grain. Visible pores, fine baby hairs at the hairline, loose strands of hair catching the light, no retouching. Avoid: plastic skin, heavy smoothing, doll-like face, dark face, grey whites.

MiniMax H3 Image: Candid portraitMiniMax H3 Image
1K · Add detail off

Dark hair, laugh, hand in hair, soft light as asked. She looks at the camera rather than away, and no mole is visible.

GPT Image 2.6: Candid portraitGPT Image 2.6
1K · economy channel

Laughing, hand in hair, looking away, with the mole. Hair is lighter brown rather than dark.

Film still: starship in the desert

A farmer and a mule for scale, hazy light, weathered materials and a 35mm film look.

More H3 Image cinematic prompts
Prompt (same for every model)

A 35mm film photograph of a colossal derelict starship half-buried in desert sand, a lone farmer leading a mule past it in the foreground for scale, hazy late-afternoon light, real-world materials and weathering, rust streaks and sand drifts against the hull, muted film colours, fine grain

MiniMax H3 Image: Film still: starship in the desertMiniMax H3 Image
1K · Add detail off

A huge hull half-buried in sand, with a man and a mule in front for scale. The hull is a plain cylinder.

GPT Image 2.6: Film still: starship in the desertGPT Image 2.6
1K · economy channel

A detailed starship half-buried in sand, with a farmer leading a mule past it. A film-frame border adds to the 35mm look.

Chinese ink landscape with poem and seal

A winter river scene in the Song style, plus a vertical poem in traditional characters and a red seal reading 寒江.

More H3 Image Chinese calligraphy prompts
Prompt (same for every model)

宋代山水畫風格,冬日寒江,大雪覆蓋遠山,江心一葉孤舟,舟上漁翁身披蓑衣、頭戴斗笠,獨自垂釣,畫面大片留白。畫面右上角豎排行書題詩「孤舟蓑笠翁,獨釣寒江雪」,題詩下方蓋一枚紅色方形印章,印文為篆書「寒江」。絹本設色,古畫質感。

MiniMax H3 Image: Chinese ink landscape with poem and sealMiniMax H3 Image
1K · Add detail off

Poem and seal text correct, in traditional characters. The snow is faint and the boat is tiny, so the heavy snow on the mountains is only loosely met.

GPT Image 2.6: Chinese ink landscape with poem and sealGPT Image 2.6
1K · economy channel

Poem and seal text correct. Snow-covered mountains and a fisherman in a hat with a rod.

Run these prompts on H3 Image

Press Try this prompt on any prompt above to load it here, or write your own.

Add photos to edit them (up to 16)

New accounts get 20 free credits · See pricing

Not affiliated with MiniMax · Your images are never used to train AI · Privacy policy