Doubao-1.5-Thinking-Vision-Pro-250428: benchmark results

Provider: ByteDance.

Aggregate Rating 1651 ± 32, rank #619 of 10916 rated models, from 15 benchmark results.

Strongest benchmark results

BenchmarkScoreMetricPercentile
OmniTraffic - Temporal Reasoning50.7Accuracy (%) on temporal reasoning (order frames and track q80
OmniTraffic - Scene Attributes85Accuracy (%) on scene attributes (weather and pedestrians in65
OmniTraffic - Multi-View Comparison85.2Accuracy (%) on multi-view comparison (compare vehicle count50
OmniTraffic - Special Event Recognition71.7Accuracy (%) on special event recognition (obstructions and 50
BABE (Biology Arena)31.71Average score (%), the paper's fixed weighting of its strong31.2
BABE (Biology Arena) - Weak Correlation31.28Score (%) on the weak-correlation triplets of independent qu31.2
OmniTraffic - Signal Phase Analysis72.8Accuracy (%) on signal phase analysis (vehicles and events p30
OmniTraffic - View-BEV Mapping26Accuracy (%) on view to bird-eye-view mapping (map between a30
BABE (Biology Arena) - Strong Correlation32.12Score (%) on the strong-correlation triplets, where each que25
OmniTraffic53.4Overall accuracy (%) over all eleven task categories, multip20
OmniTraffic - Road Infrastructure40.3Accuracy (%) on road infrastructure (lane counts and lane fu20
OSWorld33.8Success Rate (%)19.7

Interactive version: theaggregate.ai/model?slug=doubao-1-5-thinking-vision-pro-250428 · How It Works · Data refreshed daily, snapshot 2026-11-08.