192 evaluations across 6 categories
Benchmark V1 verdicts are produced by VLM judges and can contain mistakes. Treat PASS/FAIL labels as machine-assisted assessments, and inspect the images yourself. Learn more about the methodology.
Source-backed model context, size, cost, and request settings for this ImageBench V1 run.
GPT Image 2 is OpenAI's current image generation and editing model for fast, high-quality image creation, with flexible image sizes and high-fidelity image inputs through the OpenAI Images API.
GPT Image 2.5 Flare is one of two API variants OpenAI released alongside ChatGPT Images 2.5 (September 2026), positioned as OpenAI's fastest model for high-quality, everyday image generation.
On ImageBench V1 (192 prompts, 6 categories), GPT Image 2 scores 80 overall with a 87% task pass rate, while GPT Image 2.5 Flare scores 81.5 overall with a 86.5% pass rate. The biggest difference is human realism (hands, faces, anatomy): GPT Image 2 passes 69% of those prompts vs 54.8% for GPT Image 2.5 Flare. GPT Image 2 is an API model (token-based); GPT Image 2.5 Flare is an API model (token-based).
GPT Image 2.5 Flare is better overall on ImageBench V1, scoring 81.5 vs 80. GPT Image 2 is strongest at text rendering (100%). GPT Image 2.5 Flare is strongest at text rendering (100%).
Yes — on the Human realism category of ImageBench V1, GPT Image 2 passes 69% of prompts and GPT Image 2.5 Flare passes 54.8%. That is the largest per-category difference between GPT Image 2 and GPT Image 2.5 Flare.
The ImageBench gallery shows GPT Image 2 and GPT Image 2.5 Flare outputs for the same 192 prompts side by side, filterable by category and difficulty — the fastest way to judge them with your own eyes instead of trusting aggregate scores. Open them in the gallery.