Text-to-Image Generation on GenEval 2
92.6Soft TIFA GMFlow-DPPO + CPS
Evaluation results
| Method | Links | |||
|---|---|---|---|---|
| Flow-DPPO + CPSBackbone=FLUX2-9B, Training Configuration=Single2026.06 | 92.6 | — | — | |
| Flow-DPPO + CPSBackbone=FLUX.1-dev, Training Configuration=Single2026.06 | 91.6 | — | — | |
| Flow-CPSBackbone=FLUX.1-dev, Training Configuration=Single2026.06 | 91.2 | — | — | |
| Flow-DPPO + CPSBackbone=FLUX2-9B, Training Configuration=+CFG2026.06 | 91 | — | — | |
| Flow-DPPOBackbone=FLUX.1-dev, Training Configuration=Single2026.06 | 90.7 | — | — | |
| Flow-CPSBackbone=FLUX2-9B, Training Configuration=+CFG2026.06 | 89 | — | — | |
| Flow-GRPOBackbone=FLUX.1-dev, Training Configuration=Single2026.06 | 87.8 | — | — | |
| GRPO-GuardBackbone=FLUX.1-dev, Training Configuration=Single2026.06 | 87.6 | — | — | |
| Flow-DPPOBackbone=FLUX2-9B, Training Configuration=+CFG2026.06 | 87.4 | — | — | |
| GRPO-GuardBackbone=SD3.5, Training Configuration=Single2026.06 | 85.8 | — | — | |
| Flow-DPPOBackbone=FLUX2-9B, Training Configuration=Single2026.06 | 85.1 | — | — | |
| Flow-GRPOBackbone=FLUX2-9B, Training Configuration=Single2026.06 | 84.5 | — | — | |
| Flow-DPPO + CPSBackbone=SD3.5, Training Configuration=Single2026.06 | 84.1 | — | — | |
| GRPO-GuardBackbone=FLUX2-9B, Training Configuration=Single2026.06 | 82.8 | — | — | |
| Flow-CPSBackbone=FLUX2-9B, Training Configuration=Single2026.06 | 82.7 | — | — | |
| Flow-DPPOBackbone=SD3.5, Training Configuration=Single2026.06 | 78.9 | — | — | |
| GRPO-GuardBackbone=FLUX2-9B, Training Configuration=+CFG2026.06 | 78.8 | — | — | |
| Flow-CPSBackbone=SD3.5, Training Configuration=Single2026.06 | 74.8 | — | — | |
| Diffusion-NFTBackbone=SD3.5, Training Configuration=Single2026.06 | 64.5 | — | — | |
| COMFYCLAWAgent=claude-sonnet-4.5, Image Model=z-image-turbo2026.07 | 62.01 | — | — | |
| COMFYCLAWAgent=Qwen-3.6-35B-A3B, Image Model=z-image-turbo2026.07 | 58.24 | — | — | |
| Flow-DPPOBackbone=FLUX2-9B, Training Configuration=Multi2026.06 | 57.7 | — | — | |
| Flow-GRPOBackbone=SD3.5, Training Configuration=Single2026.06 | 56.6 | — | — | |
| Flow-DPPO + CPSBackbone=FLUX2-9B, Training Configuration=Multi2026.06 | 55.2 | — | — | |
| COMFYCLAWAgent=claude-sonnet-4.5, Image Model=longcat2026.07 | 54.76 | — | — | |
| Flow-GRPOBackbone=FLUX2-9B, Training Configuration=+CFG2026.06 | 54.6 | — | — | |
| Flow-DPPO + CPSBackbone=SD3.5, Training Configuration=Multi2026.06 | 51.6 | — | — | |
| GRPO-GuardBackbone=FLUX2-9B, Training Configuration=Multi2026.06 | 49 | — | — | |
| Flow-DPPOBackbone=SD3.5, Training Configuration=Multi2026.06 | 48.1 | — | — | |
| GRPO-GuardBackbone=SD3.5, Training Configuration=Multi2026.06 | 47.8 | — | — | |
| Diffusion-NFTBackbone=FLUX2-9B, Training Configuration=Multi2026.06 | 47.3 | — | — | |
| Flow-CPSBackbone=FLUX2-9B, Training Configuration=Multi2026.06 | 47.1 | — | — | |
| COMFYCLAWAgent=Qwen-3.6-35B-A3B, Image Model=longcat2026.07 | 47.08 | — | — | |
| Flow-GRPOBackbone=FLUX2-9B, Training Configuration=Multi2026.06 | 46.8 | — | — | |
| Flow-CPSBackbone=SD3.5, Training Configuration=Multi2026.06 | 44.6 | — | — | |
| Diffusion-NFTBackbone=SD3.5, Training Configuration=Multi2026.06 | 42.5 | — | — | |
| BASEAgent=claude-sonnet-4.5, Image Model=z-image-turbo2026.07 | 41.88 | — | — | |
| Flow-GRPOBackbone=SD3.5, Training Configuration=Multi2026.06 | 39.9 | — | — | |
| BASEAgent=Qwen-3.6-35B-A3B, Image Model=z-image-turbo2026.07 | 36.53 | — | — | |
| COMFYCLAWAgent=Gemma-4-E4B-it, Image Model=z-image-turbo2026.07 | 36.21 | — | — | |
| CMOModel category=Non-Autoregressive Models2026.03 | 34 | 80 | — | |
| BASEAgent=Gemma-4-E4B-it, Image Model=z-image-turbo2026.07 | 32.76 | — | — | |
| Qwen-ImageModel category=Non-Autoregressive Models2026.03 | 31.4 | 80 | — | |
| BASEAgent=claude-sonnet-4.5, Image Model=longcat2026.07 | 31.31 | — | — | |
| LatentUMVis-Gen2026.04 | 31.3 | 72.9 | — | |
| BASEAgent=Qwen-3.6-35B-A3B, Image Model=longcat2026.07 | 30.4 | — | — | |
| BagelChain of Thought (CoT)=true2026.04 | 23.1 | 70.9 | — | |
| PrefGRPOModel category=Non-Autoregressive Models2026.03 | 23 | 70.3 | — | |
| COMFYCLAWAgent=Gemma-4-E4B-it, Image Model=longcat2026.07 | 21.94 | — | — | |
| FlowGRPOModel category=Non-Autoregressive Models2026.03 | 21.8 | 70.9 | — | |
| LatentUMBase2026.04 | 21.4 | 68.1 | — | |
| FLUX.1-devModel category=Non-Autoregressive Models2026.03 | 17.1 | 64.1 | — | |
| BASEAgent=Gemma-4-E4B-it, Image Model=longcat2026.07 | 16.41 | — | — | |
| HermesFlowModel category=Autoregressive Models2026.03 | 16.1 | 63.5 | — | |
| Show-oModel category=Autoregressive Models2026.03 | 16 | 63.4 | — | |
| SD3.5-MModel category=Non-Autoregressive Models2026.03 | 15.8 | 63.7 | — | |
| Janus Pro2026.04 | 14.5 | 58.3 | — | |
| BLIP3o 8BParameters=8B2026.04 | 13.3 | 59.4 | — | |
| JanusModel category=Autoregressive Models2026.03 | 11 | 52.9 | — | |
| SDXLModel category=Non-Autoregressive Models2026.03 | 9.1 | 50.1 | — | |
| IterCompModel category=Non-Autoregressive Models2026.03 | 8.5 | 53 | — | |
| SD 2.1Model category=Non-Autoregressive Models2026.03 | 5.2 | 40.8 | — | |
| BAGELModel Type=Unified, Evaluation Protocol=evaluation with self-CoT2026.05 | — | — | 71.7 | |
| BAGEL + UNOModel Type=Unified, Evaluation Protocol=evaluation with self-CoT2026.05 | — | — | 75.1 | |
| BLIP3-o-8BModel Type=Unified, Evaluation Source=locally evaluated2026.05 | — | — | 59.8 | |
| Emu3-8BModel Type=Unified, Evaluation Source=locally evaluated2026.05 | — | — | 41.2 | |
| Flux2026.04 | — | — | 34.59 | |
| FLUX.1-devModel Type=Gen. Only2026.05 | — | — | 67 | |
| FLUX.2 4B2026.05 | — | — | 22.07 | |
| FLUX.2 4BNFE=42026.05 | — | — | 24.03 | |
| FLUX.2 4B w/ RTDMDNFE=42026.05 | — | — | 27.55 | |
| FLUX.2 9B2026.05 | — | — | 35.57 | |
| FLUX.2 9BNFE=42026.05 | — | — | 35.7 | |
| InfinityModel Type=Gen. Only, Evaluation Source=locally evaluated2026.05 | — | — | 61.2 | |
| JanusModel Type=Unified, Evaluation Source=locally evaluated2026.05 | — | — | 53.7 | |
| Janus-FlowModel Type=Unified, Evaluation Source=locally evaluated2026.05 | — | — | 42.7 | |
| Janus-Pro-7BModel Type=Unified, Evaluation Source=locally evaluated2026.05 | — | — | 54.5 | |
| Omni2026.04 | — | — | 54.12 | |
| OmniGen2Model Type=Gen. Only, Evaluation Source=locally evaluated2026.05 | — | — | 67.4 | |
| OneCAT-3BModel Type=Unified, Evaluation Source=locally evaluated2026.05 | — | — | 60.7 | |
| Qwen-Image2026.04 | — | — | 30.67 | |
| SD-3.5-LargeModel Type=Gen. Only2026.05 | — | — | 67.8 | |
| SD-3.5-MediumModel Type=Gen. Only, Evaluation Source=locally evaluated2026.05 | — | — | 64.8 | |
| SDXLModel Type=Gen. Only2026.05 | — | — | 45.3 | |
| UniWorld-V1Model Type=Unified, Evaluation Source=locally evaluated2026.05 | — | — | 70 | |
| Z-Image2026.04 | — | — | 41.83 | |
| Z-Image 6B2026.05 | — | — | 30.12 | |
| Z-Image 6B w/ TDM-R1NFE=42026.05 | — | — | 40.73 | |
| Z-Image-Turbo 6BNFE=42026.05 | — | — | 35.3 |