Benchmark

SciCode

Scientific code generation graded by test execution; subproblem context matters.

Released
2024-07-18
Categories
science
Openness
unknown
Source
model_reports
Reported scores
2

Reported scores

model_reports

ModelOrganizationReported valueReportedEvidence
Gemini 3.1 ProGoogle59.02026-02-19evidence
Kimi K3Moonshot AI58.72026-06-13evidence

Scores are partitioned by the source that reported them and are never merged into a single cross-source ranking, because the sources measure different things and say so.