Add SubQ Snap benchmark results

#42
Perch Semantic Linter org

Score all three public and private benchmarks using native SubQ decisions and Perch context sizing for the server’s 8,192-token limit. Include input-token costs, the reproducible scorer, public predictions and updated charts.

jfbrown95 changed pull request status to merged

Sign up or log in to comment