Text-to-Image Generation on GenEval 2 (Compositional and Efficiency Metrics)
1Execution Time (s)FLUX.2-dev
Evaluation Results
| Method | Links | |||||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| FLUX.2-devTemperature=0, Hardware=RTX 6000 Ada Generation 48GB GPU2026.05 | 1 | 42.09 | 83.34 | 96.04 | 95.12 | 65.52 | 73.81 | 51.06 | — | — | — | — | 66.1 | |
| FLUX.2-Klein-9BTemperature=0, Hardware=RTX 6000 Ada Generation 48GB GPU2026.05 | 1 | 34.16 | 79.06 | 95.82 | 88.86 | 57.55 | 75.13 | 37.55 | — | — | — | — | 4.31 | |
| T2I-CopilotTemperature=0, Hardware=RTX 6000 Ada Generation 48GB GPU, Base Model=FLUX.2-Klein-9B, Scaling Type=Inference-time scaling2026.05 | 1.24 | 36.56 | 81.19 | 95.91 | 93.29 | 59.25 | 77.84 | 58.49 | 4.48 | 12.61 | 2.59 | 1.24 | 127.3 | |
| EPICTemperature=0, Hardware=RTX 6000 Ada Generation 48GB GPU, Base Model=FLUX.2-Klein-9B, Scaling Type=Inference-time scaling2026.05 | 13.6 | 71.46 | 90.25 | 96.7 | 92.15 | 85.03 | 81.13 | 41.58 | 2.53 | 8.08 | 1.12 | 0.36 | 175.6 | |
| RAISETemperature=0, Hardware=RTX 6000 Ada Generation 48GB GPU, Base Model=FLUX.2-Klein-9B, Scaling Type=Inference-time scaling2026.05 | 19.8 | 52.23 | 86.28 | 97.18 | 95.59 | 68.47 | 86.57 | 46.42 | 9.12 | 38.6 | 10.78 | 7.12 | 718.4 | |
| BoN+NVILATemperature=0, Hardware=RTX 6000 Ada Generation 48GB GPU, Base Model=FLUX.2-Klein-9B, Scaling Type=Inference-time scaling2026.05 | 32 | 41.6 | 82.38 | 97.1 | 91.1 | 61.55 | 80.39 | 44.83 | — | — | — | — | 153.1 |