Unverified50% confidenceBenchmarkExact time
在综合基准测试中，So得55分（约78.6%），Terra得44分（约62.9%），Luna得31分（约44.3%）
1
Sources
50%
Confidence
Medium-term (~90 days)
Relevance
7/10/2026
First Seen
Valid until: 10/8/2026
Sources
GPT-5.6三款模型实测：So/Terra/Luna性价比全面解析
bilibili宙流AI小工坊主7/9/2026
Related Claims
Unverified在自研基准PinBash 3测试中，SOL约得78.6%，TERRA约62.9%，LUNA约44.3%76% similarPartially Verified在KingBench 3测评中Soul得55分（约78.6%），Tara得44分（约62.9%），Luna得31分（约44.3%）71% similarUnverifiedClaude Opus 4.7中间检查点通过率达57%，最终准确率却只有39.8%71% similarUnverified在贴吧小站项目评测中，Sol以83.95分居首，GLM 5.2以76.59分位列第二，Terra 75.97分排第三，Luna 68.56分排第四66% similarUnverified引导模式下平均端到端准确率从28.6%提升到33.7%，歧义检测F1值从45.3%提升到64.9%65% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/471364API
curl https://kongchang.com/api/v1/knowledge/claims/471364MCP
get_claim(id=471364)