随著美国两大AI实验室与低成本中国竞争对手之间的价格战加剧，Anthropic与OpenAI双双推出领先旗舰模型的降价版本。Anthropic推出了使用成本较前代降低约40%的Claude Opus 5.5，而OpenAI则在月初发布顶级ChatGPT-6 Astra模型后，接连推出费用减半的Sol与Luna模型，显示竞争焦点已从纯粹追求顶尖效能转向性价比的全面抗衡。

面对开源权重模型（特别是来自中国）的市占侵蚀，两大业者采取了不同的定价与市场策略以争取重视成本的企业用户。Anthropic此举旨在秋季登场的史上最大规模IPO前扩大客户群并展现增长持续性；而预计2027年以前不上市的OpenAI，则借由大幅降价刺激增长，并持续宣传其Astra模型在市场上的强大效能领导地位。

在大幅降低推论成本的同时，两家公司亦强化了安全测试与独立评估机制的布局。Opus 5.5在资安与生物等高风险领域部署了外部防护与对齐测试，体现了执行长Dario Amodei放缓前沿步伐以控制风险的呼吁；OpenAI亦公布了与第三方评估机构合作的优先事项，回应业界对超级智慧无序扩张引发安全危机的广泛担忧。

As the price war between leading US AI labs and low-cost Chinese rivals intensifies, Anthropic and OpenAI have both launched cheaper versions of their flagship models. Anthropic unveiled Claude Opus 5.5, which is roughly 40 percent cheaper than its predecessor, while OpenAI introduced lower-cost ChatGPT-6 variants, Sol and Luna, at half the price of earlier models, signaling a broader industry shift toward competing aggressively on both price and performance.

Facing growing pressure from open-weight models—particularly from China—eroding their market share, both companies are leveraging pricing strategies to appeal to cost-conscious enterprise clients. Anthropic aims to expand its enterprise user base ahead of a highly anticipated, record-breaking IPO this autumn, whereas OpenAI, not expected to go public before 2027, has slashed prices to jump-start growth following a sluggish start to the year while touting the technical superiority of its Astra model.

Alongside these aggressive price reductions, both firms are placing greater emphasis on safety guardrails and independent evaluation protocols. Anthropic's Opus 5.5 underwent rigorous alignment and external safety vetting in critical domains such as cybersecurity and biosecurity, reflecting CEO Dario Amodei's call to pace frontier development responsibly; OpenAI similarly outlined its collaboration priorities with independent evaluators amid persistent warnings about unchecked AI risks.