Unverified50% confidenceBenchmarkExact time
Opus 5在AI Omniscience测试拿了31分，避免瞎编的能力优于Opus 4.8
1
Sources
50%
Confidence
Medium-term (~90 days)
Relevance
8/23/2026
First Seen
Valid until: 11/21/2026
Sources
Related Entities
Related Claims
UnverifiedClaude Opus 4.8在AI Coding基准测试中表现超越GPT-5.576% similarUnverifiedOpus 4.8在处理高难度任务时表现出比Opus 4.7更强的持久性，不易中途放弃或陷入循环68% similarUnverifiedClaude Opus 4.8放过自己代码中漏洞的概率比4.7降低了约4倍68% similarUnverifiedClaude Opus 4.5在Anthropic内部两小时限时工程考试中得分超过了该公司历史上评估过的所有人类候选人68% similarUnverifiedOpus 4.8 的小说创作能力被评价为与 Opus 4.6 差不多，没有明显提升68% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/788651API
curl https://kongchang.com/api/v1/knowledge/claims/788651MCP
get_claim(id=788651)