Text-to-Image Generation on GenEval (Compositional Attributes and Latency)
96.97Two Objects ScoreInfinity-8B
Evaluation results
| Method | Links | ||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|
| Infinity-8BModel Scale=8B, Batch Size=1, GPU Type=NVIDIA L40 GPU2026.02 | 96.97 | — | 61 | — | 75 | — | 83 | 4.86 | 1 | — | |
| Playground v32025.12 | 95 | 99 | 72 | 82 | 50 | 54 | 76 | — | — | — | |
| Infinity-8B + ToProVARModel Scale=8B, Batch Size=1, GPU Type=NVIDIA L40 GPU2026.02 | 94.95 | — | 61 | — | 76 | — | 83 | 1.78 | 2.7 | — | |
| Infinity-8B + SkipVARModel Scale=8B, Batch Size=1, GPU Type=NVIDIA L40 GPU2026.02 | 94.94 | — | 57.5 | — | 76.5 | — | 82 | 2.11 | 2.3 | — | |
| Infinity-8B + FastVARModel Scale=8B, Batch Size=1, GPU Type=NVIDIA L40 GPU2026.02 | 94.19 | — | 57 | — | 75.25 | — | 81 | 2.01 | 2.4 | — | |
| SD3Parms=8B2025.12 | 94 | 99 | 72 | 89 | 33 | 60 | 74 | 23.3 | — | — | |
| BAGELParms=14B2025.12 | 94 | 99 | 64 | 81 | 88 | 63 | 82 | 45.1 | — | — | |
| Sparse-LaViDaParms=10.4B2025.12 | 93 | 99 | 63 | 61 | 88 | 64 | 78 | 10.86 | 1.95 | — | |
| RoPEResolution=5122025.10 | 90.4 | 98.8 | 40.8 | 84.4 | 84.3 | 60.6 | 76.5 | — | — | — | |
| RoPEResolution=10242025.10 | 90.2 | 99.1 | 37.3 | 80 | 84.8 | 63.5 | 75.7 | — | — | — | |
| HARoPEResolution=10242025.10 | 90.2 | 99.1 | 39.5 | 82.8 | 84.8 | 65.8 | 77.1 | — | — | — | |
| HARoPEResolution=5122025.10 | 89.4 | 98.8 | 38.8 | 86.6 | 82.2 | 59.3 | 75.8 | — | — | — | |
| RoPEResolution=20482025.10 | 88.9 | 97.8 | 37 | 75.3 | 80.9 | 62.3 | 73.5 | — | — | — | |
| HARoPEResolution=20482025.10 | 88.1 | 98.4 | 45.5 | 73.4 | 82.2 | 66 | 75.6 | — | — | — | |
| DALLE 32025.12 | 87 | 96 | 47 | 83 | 43 | 45 | 67 | — | — | — | |
| Flux-DevParms=12B2025.12 | 85 | 99 | 74 | 79 | 21 | 48 | 68 | 31.6 | — | — | |
| LaViDa-OParms=10.4B2025.12 | 85 | 99 | 65 | 71 | 86 | 58 | 77 | 21.27 | 1 | — | |
| Show-oParms=1B2025.12 | 80 | 98 | 31 | 66 | 84 | 50 | 68 | — | — | — | |
| Infinity-2BModel Scale=2B, Batch Size=1, GPU Type=NVIDIA L40 GPU2026.02 | 79.01 | — | 24 | — | 58 | — | 69 | 2.1 | 1 | — | |
| Infinity-2B + ToProVARModel Scale=2B, Batch Size=1, GPU Type=NVIDIA L40 GPU2026.02 | 78.8 | — | 29.5 | — | 62 | — | 69 | 0.61 | 3.4 | — | |
| Infinity-2B + FastVARModel Scale=2B, Batch Size=1, GPU Type=NVIDIA L40 GPU2026.02 | 78.79 | — | 27.75 | — | 59.5 | — | 68 | 0.8 | 2.6 | — | |
| Infinity-2B + SkipVARModel Scale=2B, Batch Size=1, GPU Type=NVIDIA L40 GPU2026.02 | 76.77 | — | 26.5 | — | 57.5 | — | 67 | 1.1 | 2 | — | |
| MMaDaParms=8B2025.12 | 76 | 99 | 20 | 61 | 84 | 37 | 63 | — | — | — | |
| SDXLParms=2.6B2025.12 | 74 | 98 | 39 | 85 | 15 | 23 | 55 | 5.2 | — | — | |
| Ambient-α (L0)evaluation_engine=GPT-4o, loops=02026.01 | 40 | 97 | 11 | 36 | 82 | — | 47 | — | — | 14 | |
| Ambient Loops (L1)evaluation_engine=GPT-4o, loops=12026.01 | 38 | 97 | 11 | 35 | 78 | — | 47 | — | — | 19 | |
| Micro-diffusionevaluation_engine=GPT-4o2026.01 | 33 | 97 | 6 | 35 | 82 | — | 44 | — | — | 14 |