List your company

Multimodal Agent Performance on CV-Bench

83.7AccuracyToolGate
77.35679.00380.6582.297Jun 2, 2026

Evaluation results

MethodLinks
ToolGateBackbone=235B-FP8, Training regime=Cross-domain2026.0683.7
BaselineBackbone=235B-FP8, Training regime=None2026.0682.7
ToolGateBackbone=30B, Training regime=In-domain2026.0680
BaselineBackbone=30B, Training regime=None2026.0678.8
ToolGateBackbone=30B, Training regime=Cross-domain2026.0677.6