GPT Image 1.5 vs 2: Which One to Use

OpenAI ships GPT Image updates fast, and Apiframe currently exposes two current model IDs: gpt-image-1.5 and gpt-image-2. Worth noting upfront: OpenAI's original GPT Image 1, released April 2025, isn't carried on Apiframe at all, only the two current versions are. Here's what actually changed between them, and which one to call.

The lineup at a glance

The real differences, straight from Apiframe's own docs

Apiframe's gpt-image-2 docs page has a dedicated section comparing it directly against gpt-image-1.5, an unusually convenient primary source. Four things actually changed:

No transparent background on GPT Image 2. gpt-image-1.5 accepts background: "transparent". gpt-image-2 only accepts auto or opaque. If you need transparent PNGs, product cutouts, logos, UI assets meant to sit on top of other content, stay on 1.5.

No input_fidelity knob on GPT Image 2. GPT Image 1.5 lets you choose between low and high fidelity when editing with a reference image. GPT Image 2 always processes input images at high fidelity, there's no lower-cost, lower-fidelity option to fall back on.

New moderation field on GPT Image 2. Not present on 1.5. Set it to low for less restrictive content filtering, auto is the default on both models but only 2 lets you loosen it.

Default output format changed. GPT Image 1.5 defaults to png. GPT Image 2 defaults to webp. Set output_format explicitly if your pipeline expects one or the other.

Pricing: nearly identical, not a premium upgrade

GPT Image 2 is marginally cheaper at high and auto quality, 23 credits versus 24, despite adding a reasoning step no earlier version had. This isn't a rounding artifact either: OpenAI's own underlying token rates put GPT Image 2's output-token price at $30 per million tokens, versus $32 per million for GPT Image 1.5, a real, if small, per-token reduction that shows up in Apiframe's credit costs too. Worth stating plainly: this is a meaningfully better model at the same or slightly lower cost, not a premium tier you pay more to access.

Which one should you actually use?

If your workflow depends on transparent backgrounds, product cutouts, sticker-style assets, UI icons, GPT Image 1.5 is the only one of the two that supports it, upgrading to 2 means losing that capability entirely. For everything else, reasoning-driven layout, multi-image consistency, stronger multilingual text, better instruction following, GPT Image 2 is the better default for new projects, and it costs the same or slightly less doing it.

Both run through the same API key and billing, so switching is a one-line change:

import requests

def generate(model, prompt, params):
    return requests.post(
        "https://api.apiframe.ai/v2/images/generate",
        headers={"X-API-Key": "afk_your_api_key_here"},
        json={
            "model": model,
            "prompt": prompt,
            ("gptImageParams" if model == "gpt-image-1.5" else "gptImage2Params"): params,
        },
    )

# Need a transparent PNG for a UI asset
icon = generate("gpt-image-1.5", "a minimalist app icon of a mountain, flat design",
                 {"background": "transparent", "quality": "high"})

# Everything else, default to the newer model
poster = generate("gpt-image-2", "a bilingual event poster in English and Japanese, clean layout",
                   {"quality": "high", "aspect_ratio": "3:2"})

No separate OpenAI account, no re-integration, just a different model string and the matching params key.

Get an API key and start with free credits, or try either model directly in Apiframe Studio without writing code. Full parameter references live in the docs, and current plan limits are on the pricing page. If you're weighing Apiframe against other ways to call GPT Image 2, see Best GPT Image 2 API Providers Compared.

FAQ

Is GPT Image 1.5 obsolete now that 2 exists?

No. It's the only one of the two that supports transparent backgrounds and an adjustable input-fidelity setting. If either matters to your workflow, stay on 1.5.

Does GPT Image 2 cost more than 1.5?

No, it's actually marginally cheaper at high and auto quality (23 credits versus 24), despite adding reasoning capability neither earlier version had.

Can GPT Image 2 generate transparent PNGs?

No. Its background parameter only accepts auto or opaque. Use gpt-image-1.5 for transparent output.

What happened to the original GPT Image 1?

OpenAI released it in April 2025. It's since been superseded by 1.5 and then 2, and isn't available as a separate model ID on Apiframe.

Do both versions support image editing?

Yes, both accept input_images for editing and combining references. GPT Image 1.5 additionally lets you set fidelity for that editing; GPT Image 2 always uses high fidelity.

Which GPT Image version has better text rendering?

GPT Image 2. It's a specifically highlighted improvement over 1.5, with stronger multilingual and dense-text accuracy.