AI benchmark · live rankings
Bro Frontier Index v2.3
More reliable and less exploitable tasks, broader computer-use and science coverage, stronger cybersecurity weighting, and fairer category weights.
Models
—
model-effort combinations
Frontier
—
—
Best open
—
—
As of
Live
latest releases
Model performance
Scores, coverage, and cost in one grid.
50+30–49<30
About these scores
Preliminary scores renormalize over measured categories.Estimated entries are tagged.
More analysis Task weights
Task weights
How categories combine.
Legacy benchmark
Bro Capabilities Index
The original seven-category capability benchmark, kept for historical comparison.
Models
—
parsed entries
Frontier
—
—
Best open
—
—
As of
Live
latest releases
BCI archive
Overall plus seven category averages.
Security benchmark
Bro Cyber Eval
Security evaluation over CY1 to CY4.
Models
—
measured
Frontier
—
—
Best open
—
—
As of
Live
latest releases
Cyber performance
Task scores and the weighted overall.
Model against model
Compare models
Line up four models across every metric.
Benchmark
BFI
active view
Models
—
side by side
Benchmark
Model details
Every metric side by side. The best value in each row is marked.