OpenAI drops GPT Image 1.5 with 4x faster speeds and precise edits

Anyone who has tried editing an AI image knows the pain of asking for a different jacket and watching the model regenerate an entirely new face and background.

OpenAI just launched GPT Image 1.5, its new flagship image generation model. It brings a 4x boost in generation speed, slashes API costs by 20%, and finally introduces precise local editing that actually listens to natural language instructions.

Local edits replace total rerolls

Built on OpenAI's GPT-5.2 multimodal architecture, GPT Image 1.5 blends a unified Transformer backbone with diffusion generation.

Generation is noticeably snappier. Simple prompts pop up in seconds, and complex requests finish in under two minutes, paired with a 30% jump in complex instruction following.

The real breakthrough is conversational editing. Instead of dealing with layers, masks, or Photoshop, users can simply say make the jacket blue or change the lighting to sunset.

The model modifies only the requested parts while preserving composition, lighting, facial identities, and brand logos across iterations. That leap helped deliver a 245-point bump in image editing performance over the original GPT Image 1.

Typography has also seen a serious upgrade. The model can accurately render small, dense text on posters and menus, maintaining low error rates on multi-line text blocks up to 800 characters. It also supports uploading up to 16 reference images for composition and style guidance.

Back on top of the arena

The upgrades immediately showed up on the LMArena leaderboard. GPT Image 1.5 claimed the number one ranking across both text-to-image with a 1264 score and image editing with a 1409 score.

Community reactions rolled in fast, with creators declaring that the king is back and noting that OpenAI has built serious competition for Nano Banana Pro.

Safety-wise, it includes native C2PA metadata authentication. The trade-off is resolution: GPT Image 1.5 caps out at 1536x1024, leaving room for rivals that currently push native 4K outputs.

Why it matters

For creators and brand teams, GPT Image 1.5 tackles the most frustrating bottleneck in AI design: consistency. Keeping human likenesses and company logos intact across revisions means these tools can finally slide into real product catalogs and marketing pipelines without requiring endless manual retouching.

On the industry side, OpenAI is putting intense pressure on incumbents like Midjourney that still rely heavily on full-canvas regenerations. By combining high-speed generation with reliable text rendering and multi-turn edits, the baseline expectation for creative AI has changed.

We think this is a smart pivot by OpenAI. Rather than chasing headline-grabbing 4K pixel counts, the team focused on practical usability, text accuracy, and workflow control. Image generation is no longer just about generating a pretty one-off picture; it is about keeping what works and tweaking only what needs fixing.