DeepEyes-7B: benchmark results

Provider: DeepEyes. Released 2025-05-20. Access: Open.

Aggregate Rating 1530 ± 45, rank #1859 of 10516 rated models, from 35 benchmark results.

Strongest benchmark results

BenchmarkScoreMetricPercentile
LongShOTBench - Short Videos27.06Weighted Rubric Score (%)64.4
LongShOTBench - Core Perception23.63Weighted Rubric Score (%)60.4
LongShOTBench - Reasoning31.23Weighted Rubric Score (%)57.4
LongShOTBench - Speech26.51Weighted Rubric Score (%)57.4
LongShOTBench - Visual26.22Weighted Rubric Score (%)57.4
LongShOTBench - Medium Videos25.87Weighted Rubric Score (%)55.9
LongShOTBench27.8Weighted Rubric Score (%)55.4
LongShOTBench - Long Videos26.8Weighted Rubric Score (%)55.4
LongShOTBench - Audio24.93Weighted Rubric Score (%)53.5
LongShOTBench - Multimodal31.52Weighted Rubric Score (%)52.5
LongShOTBench - Information24.8Weighted Rubric Score (%)49.5
ReferFlow-Bench - Reset - T-Acc7.4T-Acc (%)43.5

Interactive version: theaggregate.ai/model?slug=deepeyes-7b · How It Works · Data refreshed daily, snapshot 2026-11-21.