List your company

Image Editing on GEditBench v2 (test)

1,260Instruction Following ELOGPT Image 1.5 (26-03-04)
768.08895.791,023.51,151.21Mar 30, 2026

Evaluation results

MethodLinks
GPT Image 1.5 (26-03-04)Samples=1,081, Evaluation Protocol=GPT-4o / PVC-Judge / Bradley-Terry model2026.031,260-131,149-12846-131,071-7
Nano Banana Pro (26-03-04)Samples=1,156, Evaluation Protocol=GPT-4o / PVC-Judge / Bradley-Terry model2026.031,126-131,066-91,108-111,096-6
Seedream 4.5 (26-03-11)Samples=1,190, Evaluation Protocol=GPT-4o / PVC-Judge / Bradley-Terry model2026.031,111-121,142-111,030-111,089-7
Qwen-Image-Edit-2511Samples=1,200, Evaluation Protocol=GPT-4o / PVC-Judge / Bradley-Terry model2026.031,095-101,060-11972-91,038-6
FLUX.2 [klein] 9BSamples=1,200, Evaluation Protocol=GPT-4o / PVC-Judge / Bradley-Terry model2026.031,083-131,025-111,019-101,039-6
FLUX.2 [dev] TurboSamples=1,200, Evaluation Protocol=GPT-4o / PVC-Judge / Bradley-Terry model2026.031,068-12936-101,064-111,021-6
FLUX.2 [dev]Samples=1,200, Evaluation Protocol=GPT-4o / PVC-Judge / Bradley-Terry model2026.031,037-12965-101,018-111,006-7
Qwen-Image-Edit-2509Samples=1,200, Evaluation Protocol=GPT-4o / PVC-Judge / Bradley-Terry model2026.031,033-101,062-10955-91,014-5
LongCat-Image-EditSamples=1,200, Evaluation Protocol=GPT-4o / PVC-Judge / Bradley-Terry model2026.031,018-10968-101,017-101,001-6
FLUX.2 [klein] 4BSamples=1,200, Evaluation Protocol=GPT-4o / PVC-Judge / Bradley-Terry model2026.031,007-121,019-101,070-101,031-6
Qwen-Image-EditSamples=1,200, Evaluation Protocol=GPT-4o / PVC-Judge / Bradley-Terry model2026.03991-101,073-11971-111,010-6
Step1X-Edit-v1p2Samples=1,200, Evaluation Protocol=GPT-4o / PVC-Judge / Bradley-Terry model2026.03909-121,007-121,067-11996-6
FLUX.1 Kontext [dev]Samples=1,200, Evaluation Protocol=GPT-4o / PVC-Judge / Bradley-Terry model2026.03849-13900-13840-14869-7
BagelSamples=1,200, Evaluation Protocol=GPT-4o / PVC-Judge / Bradley-Terry model2026.03820-13694-17987-13851-8
OmniGen V2Samples=1,200, Evaluation Protocol=GPT-4o / PVC-Judge / Bradley-Terry model2026.03807-13910-12929-13888-7
GLM-ImageSamples=1,200, Evaluation Protocol=GPT-4o / PVC-Judge / Bradley-Terry model2026.03787-131,023-111,109-13979-6