GPT Image 1.5: Faster, More Faithful Edits

OpenAI has released GPT Image 1.5 — a faster, more instruction-faithful image model that the newsletter reports is rolling out to ChatGPT users and available in the API as GPT-Image-1.5. According to the writeup, the model is up to 4x faster than the prior version and much better at following editing instructions, preserving lighting, composition, and people’s appearance across multiple edits.

The piece lists concrete strengths: precise editing (add/subtract/combine elements without breaking the image), improved text rendering for dense/small text, stronger instruction following (accurate object placement and complex compositions), creative style transforms, and major speed gains. OpenAI also shipped a dedicated Images interface in the ChatGPT sidebar with preset filters, trending prompts, and one-time likeness uploads for easier creation.

Benchmarks cited in the article show GPT Image 1.5 leading the Artificial Analysis text-to-image leaderboard with 1,264 points (a 29‑point lead) and topping LM Arena’s image-editing ranking at 1,409 points — narrowly beating Google’s Nano Banana Pro (2K) by 3 points. Reported improvements versus GPT Image 1.0 include +147 points in text-to-image and +245 in editing on LM Arena. The newsletter frames the tradeoff: GPT Image 1.5 = speed + precise edits, while Nano Banana Pro targets production-grade assets and enterprise features.

If you produce visual content, the article recommends trying the new ChatGPT Images interface or running head‑to‑head tests in LM Arena / AA Image Arena to see which model fits your workflow.

Source