ModelsMulti-source79°

GPT-6.2 Astra 在 Agents' Last Exam 1.1 上以 5.5 分领先 Claude Fable 5.2

Why we picked it

如果你们团队在跑 Agent 任务,这条对比直接关系账单:GPT-6.2 Astra 分数高 5.5 分,价格只有 Claude Fable 5.2 的 5%。做模型选型前可以先看看这份评测细节。

OpenAI 的 GPT-6.2 Astra 在基准测试 Agents' Last Exam 1.1 上超过 Anthropic 的 Claude Fable 5.2,领先幅度为 5.5 分。消息通过 X 平台发布,核心卖点是成本差距:GPT-6.2 Astra 的价格仅为 Claude Fable 5.2 的 5%。

Body

这条对比强调的是智能密度与成本的组合:同样的智能体评测分数下,价格相差约 20 倍。对于按 token 付费的开发者和企业用户来说,跑大规模 Agent 工作流时,模型单价的差距会直接放大成总成本的差距。不过该数据来自单条推文,测试分数的具体配置、任务覆盖范围和复现细节尚未公布,实际效果还需参考完整评测报告后再决定是否迁移现有工作流。

Original · Yuchen Jin

GPT-6.2 Astra beats Claude Fable 5.2 by 5.5 points on Agents' Last Exam 1.1, at just 5% of the cost.

Improve the intelligence-per-dollar! https://t.co/VMyK8zVLiy

Industry Coordinates & Source Verification

This report is categorized under AITOP's "Models" desk, focusing on US frontier AI labs and Silicon Valley technology leaders. Within our cross-regional analytical framework, it directly maps to US-China foundation model benchmarks and commercialization comparison. Key entities and technical topics under continuous tracking include GPT-6.2, Claude, OpenAI, Anthropic.

Source attribution: curated from Yuchen Jin (@Yuchenj_UW). Cross-source corroboration: 11 independent sources have covered this development (including Artificial Analysis, IT之家, lmarena.ai), confirming industry-wide signal strength.