Unverified50% confidenceBenchmarkExact time
Scanning thresholds on Sonnet 5.5 found that above 0.88 cached answers no longer show a quality disadvantage while still saving 64% of cost; 0.90 saves 61%.
1
Sources
50%
Confidence
Medium-term (~90 days)
Relevance
10/19/2026
First Seen
Valid until: 1/17/2027
Sources
Related Entities
Related Claims
UnverifiedSonnet 5.5 uses about 60% more output tokens than Opus 5.5 to complete tasks at the max effort setting, resulting in a higher per-task cost.71% similarUnverifiedFor the same failing test fix, Haiku costs only about one-sixth of a cent—17 times cheaper than Sonnet—and completes in 6 seconds (compared to 10 seconds for Sonnet).69% similarUnverifiedThe improvement in agentic capabilities is accompanied by exponential growth in token consumption, and Sonnet 5's token consumption in maximum-effort mode is comparable to OPUS but with worse results.68% similarUnverifiedSlack's principal engineer reported that, without modifying existing prompts, Sonnet 5.5 performs better and faster, with output tokens reduced by approximately 14%.68% similarUnverifiedAt the 0.81 threshold, the proportion of cached answers that "fit the question" is 5.8 to 10.7 percentage points lower than fresh answers; converted to all responses, the overall quality loss is about 2 to 3 percentage points; Haiku 5.5 has the largest gap (10.7 percentage points), with a cached-answer pass rate of only 67.9%.67% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/1065410API
curl https://kongchang.com/api/v1/knowledge/claims/1065410MCP
get_claim(id=1065410)