List your company

Real-world Visual Understanding on RealWorldQA (test)

77.1Final PerformancePerceptionLM-8B
53.90859.92965.9571.971Apr 6, 2026Apr 17, 2026Apr 28, 2026May 9, 2026May 20, 2026May 31, 2026Jun 11, 2026

Evaluation results

MethodLinks
PerceptionLM-8BParameter Scale=8B2026.0577.1———
Molmo2-4BParameter Scale=4B2026.0573.8———
PerceptionLM-3BParameter Scale=3B2026.0573.1———
BAGEL# Params=14B2026.0672.8———
Qwen3-VL-8BParameter Scale=8B2026.0571.9———
Molmo2-8BParameter Scale=8B2026.0571.8———
Qwen3-VL-4BParameter Scale=4B2026.0571———
BLIP3-o# Params=8B2026.0669———
Full RTBackbone=Qwen3-VL-8B+2B-R2026.0468.9———
HYDRA-X# Params=7B2026.0668.7———
PerceptionLM-1BParameter Scale=1B2026.0568.6———
MAIBackbone=Qwen3-VL-2B/8B2026.0468.4509105151
Qwen2.5-VL# Params=7B2026.0668.4———
InternVL3.5-4BParameter Scale=4B2026.0567.8———
8B(S) BaselineBackbone=Qwen3-VL-8B, Mode=Standard2026.0467.6———
InternVL3.5-8BParameter Scale=8B2026.0567.4———
Zamba2-VL-7BParameter Scale=7B2026.0566.5———
LLaVA-OV# Params=7B2026.0666.3———
Qwen3-VL-2BParameter Scale=2B2026.0566———
Zamba2-VL-1.2BParameter Scale=1.2B2026.0565.9———
Show-o2# Params=7B2026.0664.7———
HYDRA# Params=7B2026.0664.7———
Zamba2-VL-2.7BParameter Scale=2.7B2026.0561.7———
InternVL3.5-2BParameter Scale=2B2026.0561.6———
2B(S) BaselineBackbone=Qwen3-VL-2B, Mode=Standard2026.0460.7———
2B(R) BaselineBackbone=Qwen3-VL-2B, Mode=Reasoning2026.0460.5———
Janus-Pro# Params=7B2026.0658———
InternVL3.5-1BParameter Scale=1B2026.0556.9———
TokenFlow-XL# Params=14B2026.0656.6———
LLaVA-1.5# Params=7B2026.0654.8———