Expired85% confidenceFactTime unknown
The test chose Claude Sonnet 4.6 rather than Opus 4.7 as the baseline, on the grounds that Opus is too strong to provide a meaningful comparison.
1
Sources
85%
Confidence
Medium-term (~90 days)
Relevance
5/31/2026
First Seen
This claim has expired. It is kept for the record and no longer describes the current state.
Valid until: 8/29/2026(expired)
Sources
Related Entities
Related Claims
UnverifiedThe evaluation chose Claude 4 Sonnet rather than Claude 4.1 Opus as the comparison subject because most developers would not use Opus in real-world scenarios due to cost considerations.80% similarUnverifiedCompared to Opus 4.6, Sonnet 4.6 performs nearly on par on most benchmarks, but with faster inference speed and lower call costs.80% similarUnverifiedClaude Sonnet 4.6 scored higher than Opus 4.6 on the GDPVal AA benchmark.77% similarUnverifiedSonnet 4.6 performs on par with Opus on the 'needle-in-a-haystack' test.77% similarUnverifiedIn agentic search tests, Sonnet 5 outperformed Sonnet 4.6 across all effort levels, but its overall performance was below Opus 4.8, placing it at approximately the Opus 4.7 level overall.76% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/13173API
curl https://kongchang.com/api/v1/knowledge/claims/13173MCP
get_claim(id=13173)