Unverified50% confidenceBenchmarkExact time

在4个智能体编码测试中，Argon赢了两个，Frontier SWE和Terminal Bench两个测试中垫底

1

Sources

50%

Confidence

Medium-term (~90 days)

Relevance

10/3/2026

First Seen

Valid until: 1/1/2027

Sources

Gemini 4 Argon深度解析：自测领先，外部验证为何存疑

Related Entities

Gemini 4 Argonmodel Frontier SWEconcept Terminal Benchconcept

Related Claims

Unverified在四个大型项目构建测试中，Fable拿下两次明确胜利加一次小胜，一次打平62% similar Unverified在VALS的Vibe Code Bench编码测试中，Claude Sonnet 5.5以92.4对91.9险胜Argon62% similar UnverifiedArgon在评估漏洞修复能力的CWE-bench v1上以68%的最高分并列第一60% similar

Cite This Claim

Stable URI

https://kongchang.com/claim/969181

API

curl https://kongchang.com/api/v1/knowledge/claims/969181

MCP

get_claim(id=969181)