GPT Image 2 vs GPT Image 1.5: What Actually Changed?

TL;DR

OpenAI now positions GPT Image 2 as the current state-of-the-art image model, while GPT Image 1.5 is labeled as the previous image generation model.

The biggest practical changes visible in official docs are:

• GPT Image 2 becomes the new default image model
• the guide now centers examples on GPT Image 2
• pricing shifts by quality tier
• GPT Image 2 supports flexible image sizes in the guide and model docs

The official positioning

| Model | How OpenAI currently describes it | |---|---| | GPT Image 2 | state-of-the-art image generation model | | GPT Image 1.5 | previous image generation model |

That alone makes the product direction clear.

Pricing differences

For the guide's example sizes, the 1024×1024 comparison looks like this:

| Quality | GPT Image 2 | GPT Image 1.5 | |---|---:|---:| | Low | $0.006 | $0.009 | | Medium | $0.053 | $0.034 | | High | $0.211 | $0.133 |

So GPT Image 2 is not simply "more expensive" or "cheaper."

The better summary is:

• cheaper at low quality
• more expensive at medium and high quality

Workflow differences that matter

The image-generation guide calls out two especially important GPT Image 2 traits:

1. flexible image sizes
2. high-fidelity image inputs

That makes GPT Image 2 easier to frame as a production model for teams that care about:

• exact output sizes
• reference-image editing
• more controlled commercial workflows

What stayed similar

Both models still live in the same broader GPT Image family and both sit inside the image-generation tooling documented by OpenAI.

You still need to think about:

• prompt quality
• edit vs generation workflow
• size and quality tradeoffs
• text and image input costs

Which one should teams use?

| Team situation | Better starting point | |---|---| | you want the current flagship | GPT Image 2 | | you need older workflow continuity | GPT Image 1.5 | | you are highly cost-sensitive on low-quality drafts | test GPT Image 2 low vs your current pipeline | | you produce polished marketing assets | benchmark GPT Image 2 medium/high directly |

The practical takeaway

GPT Image 2 is not just a rename. The docs now treat it as the lead image model, and the pricing/profile changes are material enough that teams should rerun benchmarks rather than assuming GPT Image 1.5-era economics still hold.

Sources

Source: OpenAI Developers — GPT Image 2 model page Source: OpenAI Developers — GPT Image 1.5 model page Source: OpenAI Developers — Pricing Source: OpenAI Developers — Image generation guide

GPT Image News is not affiliated with OpenAI. All trademarks belong to their respective owners.