Primary sourcelmarena.ai
View sourceEvent ·Multi-source
GPT-6.2 Astra 与 Sol 登上 Code Arena WebDev 帕累托前沿
OpenAI 两款新模型在 Code Arena: WebDev 真实场景榜单拿到成绩:GPT-6.2 Astra (Max) 以 1781 分排第 3,GPT-6.2 Sol (Max) 以 1757 分排第 7,双双进入帕累托前沿。 Astra 相比上一代 GPT-6 Astra (Max) 提高 18 分,在 Consumer Product 类别从第 4 升到第 1;单任务中位成本从 $9.46 降到 $5.45,下降 42%。Sol 相比 GPT-6.1 Sol (Max) 提高 13 分,单任务中位成本从 $1.84 降到 $1.45,下降 21%。 两款模型的定位差异明显:Astra 分数比 Sol 高 24 分,但 Sol 的单任务成本只有 Astra 的 27%($1.45 对 $5.45)。官方表示 Agent 类榜单成绩即将公布。
Current read
WebDev Arena 的新成绩单出来了,想选模型写前端的朋友可以直接看:要质量选 Astra(第 3 名、Consumer Product 第 1),要省钱选 Sol($1.45 一单,比上代还便宜 21%)。两个都在性价比前沿上,对比数据很直观。
11 sources79° AI heatLast updated 2026/11/18 17:54:46
Evidence
11 sourcesRelated source 1Artificial Analysis
View sourceRelated source 2Vaibhav Srivastav
View sourceRelated source 3kimmonismus
View sourceRelated source 4@OpenAIDevs
View sourceRelated source 5Yuchen Jin
View sourceRelated source 6TestingCatalog
View sourceRelated source 7Decoder
View sourceRelated source 8shao__meng
View sourceRelated source 9Genspark
View sourceRelated source 10@koltregaskes
View sourceConflict check
Dedup data only shows that these sources discuss the same event, not that they agree. There is no structured supporting or refuting evidence yet, so no conflict is inferred automatically.