Visual Understanding on SEED-Bench
75.1SEED ScoreSPAR-3B
Evaluation Results
| Method | Links | |
|---|---|---|
| SPAR-3BLLM Params=2B, Model Category=Und. and Gen.2026.06 | 75.1 | |
| InternVL3-2BLLM Params=2B, Model Category=Und. Only2026.06 | 75 | |
| BLIP3-o-4BLLM Params=4B, Model Category=Und. and Gen.2026.06 | 73.8 | |
| Tar-7BLLM Params=7B, Model Category=Und. and Gen.2026.06 | 73 | |
| Janus-Pro-7BLLM Params=7B, Model Category=Und. and Gen.2026.06 | 72.1 | |
| DualTokenLLM=LLaMA-2-7B, Resolution=2562026.03 | 71.8 | |
| EvoTokLLM=Qwen-2.5-7B-Inst., Resolution=2562026.03 | 71.8 | |
| SPAR-1BLLM Params=1B, Model Category=Und. and Gen.2026.06 | 71.5 | |
| InternVL3-1BLLM Params=1B, Model Category=Und. Only2026.06 | 71.1 | |
| TokLIPLLM=Qwen2.5-7B-Inst., Resolution=3842026.03 | 70.4 | |
| ShareGPT4VLLM=Vicuna-7B, Resolution=3362026.03 | 69.7 | |
| MMAR-7BType=Joint Prob. Models, LLM=Qwen-2-7B, V-Token=vae, Resolution=2562024.10 | 68.63 | |
| EMU-3Type=Joint Prob. Models, LLM=8B from scratch, V-Token=vq-vae, Resolution=5122024.10 | 68.2 | |
| EMU3LLM=8B from scratch, Resolution=-2026.03 | 68.2 | |
| Emu3-Chat-8BLLM Params=8B, Model Category=Und. Only2026.06 | 68.2 | |
| LLaVA-v1.5LLM=Vicuna-1.5-13B, Resolution=3362026.03 | 68.1 | |
| Harmon-1.5BLLM Params=1.5B, Model Category=Und. and Gen.2026.06 | 67.1 | |
| SEED-X-13BLLM Params=13B, Model Category=Und. and Gen.2026.06 | 66.5 | |
| LLaVA-OVLLM Params=1B, Model Category=Und. Only2026.06 | 65.5 | |
| TokLIP-7BLLM Params=7B, Model Category=Und. and Gen.2026.06 | 65.2 | |
| SemHiTokLLM=Qwen2.5-7B, Resolution=2562026.03 | 62.9 | |
| EMU-2Type=Und. & Gen. Non-prob., LLM=LLaMA-13B, V-Token=CLIP, Resolution=4482024.10 | 62.8 | |
| TokenFlow-LLLM=Vicuna-13B, Resolution=2562026.03 | 62.6 | |
| Unified-IO 2LLM=6.8B from scratch, Resolution=3842026.03 | 61.8 | |
| VILALLM=LLaMA-2-7B, Resolution=3362026.03 | 61.1 | |
| TokenFlow-BLLM=Vicuna-13B, Resolution=2242026.03 | 60.4 | |
| VILA-ULLM=LLaMA-7B, Resolution=3842026.03 | 59 | |
| LLaVA-1.5Type=Und. Only, LLM=Vicuna-1.5-7B, V-Token=CLIP, Resolution=3362024.10 | 58.6 | |
| Qwen-VL-ChatLLM=Qwen-7B, Resolution=4482026.03 | 57.7 | |
| VILA-UType=Joint Prob. Models, LLM=LLaMA-2-7B, V-Token=vq-vae, Resolution=2562024.10 | 56.3 | |
| VILA-ULLM=LLaMA-7B, Resolution=2562026.03 | 56.3 | |
| VILA-U-7BLLM Params=7B, Model Category=Und. and Gen.2026.06 | 56.3 | |
| MMAR-0.5BType=Joint Prob. Models, LLM=Qwen-2-0.5B, V-Token=vae, Resolution=2562024.10 | 55.7 | |
| Show-oType=Joint Prob. Models, LLM=Phi-1.5B, V-Token=CLIP, Resolution=3362024.10 | 51.61 | |
| Transfusion*Type=Joint Prob. Models, LLM=Qwen-2-0.5B, V-Token=vae, Resolution=2562024.10 | 42.4 | |
| Chameleon-7BType=Joint Prob. Models, LLM=7B from scratch, V-Token=vq-vae, Resolution=5122024.10 | 34.61 |