Skip to main content
PolicyBench

Claude Sonnet 5

Anthropic · AI alone, no tools
United States rank
#21 of 23
US exact
69.4%
United States

United States benchmark

Exact match
69.4%
Bounded score
80.7%
Parse rate
100.0%
Eligibility flags
903/984

Score by programhardest first

ProgramExactWithin 1%n
Federal tax before refundable credits44.0%47.0%100
State tax before refundable credits54.0%56.0%100
SNAP69.0%69.0%100
Person-level Medicaid eligibility74.6%74.6%177
Payroll tax76.0%86.0%100
Federal refundable credits78.0%78.0%100
State refundable credits78.0%78.0%100
Person-level CHIP eligibility91.5%91.5%177
Person-level WIC eligibility94.4%94.4%177
Person-level Head Start eligibility94.7%94.7%38
Self-employment tax96.0%96.0%100
Person-level Medicare eligibility96.0%96.0%177
SSI97.0%97.0%100
Free school meals eligibility99.0%99.0%100
Local income tax99.0%99.0%100
Reduced-price school meals eligibility99.0%99.0%100
TANF99.0%99.0%100
Person-level Early Head Start eligibility100.0%100.0%38

Hardest casesworst misses on positive references

Scores are from the frozen manuscript snapshot under the AI-alone condition: one structured response per household, no tools, graded against PolicyEngine reference outputs. See the leaderboard and paper for methodology.