Benchmark finds GPT Image 2 cheaper
Konstantin Komelin updated his AI image generation benchmark to compare 22 cloud models on the same prompt, adding Black Forest Labs' FLUX 2 Klein variants and OpenAI's GPT Image 2. The report focuses on per-image cost and wall-clock latency via Vercel AI Gateway, with side-by-side outputs, charts, and implementation notes for provider-specific quirks. The standout headline is that GPT Image 2 drops to roughly $0.0237 per image from GPT Image 1's $0.16667 while staying near the same latency, and the cheapest overall options are now Prodia Flux Fast Schnell and the FLUX 2 Klein models.
The main takeaway is not that one model wins outright, but that image-generation pricing is compressing quickly enough that model choice now looks more like infrastructure tuning than brand loyalty. GPT Image 2 looks like a major OpenAI price correction rather than a speed breakthrough, while FLUX 2 Klein 4B and 9B stand out for teams optimizing batch-generation cost. Because the benchmark captures real API-facing behavior and provider-specific parameter quirks, it is useful operationally, but since every model was tested on one prompt, it should be read as a cost-and-latency snapshot rather than a definitive quality ranking.
DISCOVERED
45d ago
2026-04-23
PUBLISHED
45d ago
2026-04-23
RELEVANCE
AUTHOR
kkomelin
