Unverified50% confidenceBenchmarkExact time
Haiku 5.5在不同推理强度之间的能力跃升明显，从high到max/X-high有相当大的提升幅度，不像Opus 5.5或Sonnet
1
Sources
50%
Confidence
Medium-term (~90 days)
Relevance
10/8/2026
First Seen
Valid until: 1/6/2027
Sources
Related Entities
Related Claims
Unverified在max等高推理强度设置下，Haiku 5.5的token消耗显著增加，单任务成本可能比GPT-6 Luna更不划算75% similarUnverified官方表示 Claude Haiku 5.5 在表格处理上的表现相较于同价位的其他模型更好74% similarUnverifiedHaiku 5.5在Chartography基准上从6.4%提升至46.4%74% similarUnverifiedHaiku 5.5的具体性能基准、定价细节与上下文窗口等关键参数尚未在公开信息中披露73% similarUnverifiedHaiku 5.5 在计算机使用（computer use）任务上的得分从 16% 跃升至 72%72% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/989554API
curl https://kongchang.com/api/v1/knowledge/claims/989554MCP
get_claim(id=989554)