Unverified50% confidenceBenchmarkExact time

GPT-5.4准确率为50.5%，其配对的永远说涨基准为51.6%

1

Sources

50%

Confidence

Long-term

Relevance

9/11/2026

First Seen

Sources

本地Gemma对决GPT-5.4：都输给了每天说"涨"的机器人

Related Entities

Related Claims

Unverified在GPT-4发布初期，主流模型在GPQA上的准确率约为35-40%，人类领域专家基准线约为65%，随机猜测为25%77% similar Unverified在Agents' Last Exam评测中，GPT-5.6 Sol取得53.6分，比采用自适应推理的Claude Fable 5高出13.1分75% similar Unverified本地运行的Gemma 26B在453次评分预测中准确率仅为49.2%，其配对的永远说涨基准准确率为53.4%74% similar

Cite This Claim

Stable URI

https://kongchang.com/claim/903348

API

curl https://kongchang.com/api/v1/knowledge/claims/903348

MCP

get_claim(id=903348)