Models

OpenAI 发布 GPT-6.2 Astra 与 Sol:成本降至 5%,跑分超越 Claude Fable 5.2

Why we picked it

OpenAI 这次把价格打到 Claude Fable 5.2 的 5%,分数还高 5.5 分,Sol 也压过了 Opus 5.5。要是你在跑 agent 或者批量调用 API,这份新的价格表值得马上对一遍,能省不少钱。

OpenAI 正式发布 GPT-6.2 系列两款模型 Astra 和 Sol。官方披露,在 Agents' Last Exam 基准上,GPT-6.2 Astra 比 GPT-6 Astra 高出 5.2 分,成本不到前者的三分之一;对比 Anthropic 的 Claude Fable 5.2 则高出 5.5 分,而成本仅为对方的 5%。

Body

GPT-6.2 Sol 的定位更偏高强度任务,公告称其成绩超过了 Opus 5.5。核心卖点是价格与性能的比值:用 5% 的成本拿到比竞品高 5.5 分的代理能力,直接瞄准智能体批量调用场景。

对开发者而言,这一轮迭代的关键变化是代理基准分数与成本同时改善,API 大规模跑 agent 的单位成本门槛显著下降;对 Anthropic 来说,Fable 5.2 在代理赛道的价格优势被明显压缩,两家在 agent 推理定价上的竞争进一步升级。

Original · TestingCatalog

OPENAI 🔥: GPT 6.2 Astra and Sol have been officially announced!

> "On the Agents’ Last Exam benchmark, GPT‑6.2 Astra is scoring 5.2 points higher than GPT‑6 Astra at less than a third of the cost, and 5.5 points ahead of Claude Fable 5.2 at just 5% of the cost."

Based on the announced benchmarks, we are getting a much cheaper model than Fable 5.2. 6.2 Sol outperforms Opus 5.5 as well.

Testing time! 👀

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 and AI developer toolchains, coding agents, and engineering workflows. Key entities and technical topics under continuous tracking include OpenAI, GPT-6.2, Claude, agents.

Source attribution: curated from TestingCatalog (@testingcatalog).