Unverified50% confidenceBenchmarkExact time
Claude Sonnet 5.5 had a fidelity rate of 93.2% and Claude Haiku 5.5 had a fidelity rate of 94.6%, with the smaller model performing better.
1
Sources
50%
Confidence
Medium-term (~90 days)
Relevance
10/11/2026
First Seen
Valid until: 1/9/2027
Sources
Related Entities
Related Claims
VerifiedClaude Sonnet 5.5 achieved 70.6% on the Terminal benchmark, compared to only 10.3% for the previous generation.78% similarUnverifiedOn the Frontier Code benchmark, GPT-6 Luna scored 56.1%, Claude Sonnet 5.0 scored 56.2%, and Kimi K3 scored 58.2%.70% similarUnverifiedOn some dimensions, Claude Sonnet 5.5 trails the more expensive Opus 5.5 by only 2 points.70% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/1000254API
curl https://kongchang.com/api/v1/knowledge/claims/1000254MCP
get_claim(id=1000254)