Model · Previous generation

GPT Image 1.5
fast and lightweight

GPT Image 1.5 is the previous-generation text-to-image model — small, quick, and ideal for rough drafts and high-volume iteration. Lower fidelity than GPT Image 2, but unbeatable on turnaround time.

Strengths of the older model

A smaller model trades photoreal polish for speed and cost.

Fast iteration
Sub-second to ~1.5s render times. When you're rapidly trying variations of a concept, fewer seconds per render compounds into a meaningful productivity win.
Lower cost
Lower per-image inference cost than GPT Image 2 — relevant if you're processing batches of prompts or running at scale.
Stable output
Consistent style across a series of prompts. Useful when you need a coherent set of images for a storyboard, mood board, or product gallery.

Where GPT Image 1.5 falls short

If you hit any of these, switch to GPT Image 2.

Photorealism. 1.5 leans more illustrative. Skin detail, fabric weave, and complex lighting can look slightly off compared to GPT Image 2.

Text in images. Anything beyond a few characters is unreliable. For headlines, signage, or labels in the image, GPT Image 2 is the better choice.

Complex scenes. Multiple characters, intricate environments, or multi-object compositions stress the older model. The newer one handles multi-subject layout better.

Prompt nuance. Specific camera angles, lens choices, and named art styles are followed less precisely. Expect to over-describe to get the look you want.

When GPT Image 1.5 still wins

Drafts
Quick concept sketches
Mood boards, layout studies, "show me what this could look like" thumbnails before committing to a real render.
Volume
Bulk variations
Generating dozens of variations on the same idea where speed and cost matter more than per-image polish.
Style
Stylized illustration
Flat, vector, low-poly, pixel-art, hand-drawn looks where photoreal fidelity isn't the point.
Iteration
Fast feedback loops
Tight design reviews where you re-prompt every 30 seconds — sub-second turnaround keeps the conversation flowing.

Try GPT Image 1.5 in the Studio

Open the studio, type a prompt, see how it compares.

Open the Studio