subjectKind
EvalResult
statement
Anthropic's Claude Sonnet 4.5 launch announcement is the source for the 77.2% SWE-bench Verified default-compute EvalResult for Claude Sonnet 4.5.
confidence
high
provenanceKind
vendor-doc
evidenceStrength
strong
status
accepted
subjectId
attribute
score
value
0.772
evidenceSourceIds
claimedAt
2026-05-04T00:00:00Z
claimedBy
a5c-ai-catalog
claimKind
evaluation
note
Replaces the removed source comment from graph/benchmarks/eval-results/swe-bench-verified-claude-sonnet-4-5-result.yaml.