Unverified50% confidenceBenchmarkExact time

Terra和Sol模型在HarvestBench测试中表现最为仁慈（杀害率最低），而GPT-4o-mini表现最为残酷（杀害率最高）

1

Sources

50%

Confidence

Short-term (~14 days)

Relevance

9/8/2026

First Seen

Valid until: 9/22/2026

Sources

HarvestBench：首个量化AI避免伤害动物意愿的基准测试

Related Entities

Terramodel Solmodel GPT-4o-minimodel

Related Claims

Unverified综合实测结果，四款模型能力排名为：Sol > GPT-5.5 > Terra > Luna67% similar UnverifiedGPT-5.6 Sol在Humanity's Last Exam取得最高分，且价格远低于Claude Opus和Fable62% similar Unverified基准测试结果显示，Sol和Sol Ultra在大多数领域优于竞争对手Misos 5，Tara优于Fable 557% similar UnverifiedMiniMax M3 的基准测试成绩亮眼，部分指标据称可比肩 GPT-5.556% similar

Cite This Claim

Stable URI

https://kongchang.com/claim/875794

API

curl https://kongchang.com/api/v1/knowledge/claims/875794

MCP

get_claim(id=875794)