List your company

Multimodal Agent Performance on V*

80.6AccuracyBaseline
75.71276.98178.2579.519Jun 2, 2026

Evaluation results

MethodLinks
BaselineBackbone=235B-FP8, Training regime=None2026.0680.6
ToolGateBackbone=235B-FP8, Training regime=Cross-domain2026.0680.1
ToolGateBackbone=30B, Training regime=In-domain2026.0679.1
ToolGateBackbone=30B, Training regime=Cross-domain2026.0676.4
BaselineBackbone=30B, Training regime=None2026.0675.9