人工智能行业正在经历一场剧烈转变，其标志是激烈竞争、激进的降本策略，以及低价模型的快速涌现。OpenAI、Google与Anthropic等头部玩家，如今在其入门级产品上把降价置于高级特性之前，以应对Kimi K3、DeepSeek V4 Flash等中国模型的竞争。这一趋势不仅在重塑市场，也在挑战那些曾经依赖高毛利定价的传统商业模式。
成本驱动竞争的新纪元
近年AI格局已显著演化，技术进展与全球经济压力共同驱动。美国公司长期重投前沿研发，而中国开发者则借助产业规模，以极低价格交付性能相当的模型。Kimi K3与DeepSeek V4 Flash的出现是这一转变最明显的例证，它们迫使西方公司重新审视定价策略与模型能力。
例如，OpenAI近期把基座模型ChatGPT 5.6 Luna的每百万token价格下调了80%，把中档模型5.6 Terra下调20%。这只是行业整体趋势的一部分：竞争的焦点不再只是性能，而是可负担性。
从高端到普惠的迁移
竞争迫使AI开发者把“可用”置于“独占”之前。OpenAI的最新动作反映了这一转变：顶级模型仍然昂贵，但更实惠的选项明显变便宜了。GPT 5.6 Luna如今输入价0.20美元/百万token、输出1.20美元/百万token，与此前的2.50美元和15美元形成鲜明对比。这一策略让OpenAI与中国对手拉开的对价更趋一致——DeepSeek V4以约0.435美元/百万token输入、0.87美元/百万token输出提供相近能力。
与此同时，能力更强的GPT 5.6 Terra现定价为输入2.00美元、输出12.00美元每百万token，低于K3的3.00美元与15.00美元。尽管有这轮降价，OpenAI也推出了旗舰模型的加速版本，例如GPT 5.6 Sol的Fast模式，按输入10美元、输出60美元每百万token计费，用于对标Claude Fable 5与Mythos 5等旗舰产品。
财务压力与效率诉求
激进定价发生在许多AI公司面临显著财务压力的时刻。OpenAI一直在与利润下滑和营收目标未达作斗争，其降价决策部分源于在竞争加剧的市场中维持商业模式的必要。
OpenAI对云计算基础设施的依赖也变得更加关键。公司今年早些时候放弃了自建数据中心的计划，这使得与Neoclouds、Oracle等供应商的租赁协议更为重要。但这些承诺代价高昂——OpenAI已承诺到2030年在算力上投入6000亿美元，远超其当前营收增速。这种财务紧张并非个例：Google去年在AI基础设施上的支出接近其云业务营收的九倍；Anthropic则因与xAI就Colossus数据中心达成的有限协议，才开始录得利润。这些例子说明，打造更强大模型的竞赛代价极其高昂。
杰文斯悖论与AI的未来
业内广泛讨论的一个理论是杰文斯悖论（Jevons Paradox）：当技术效率提升时，总用量往往随之上升。放到AI语境中，这意味着更低的token成本可能带来更高的采用率与更大的需求。分析师认为，如果token显著变便宜，用户会更倾向于大量使用，从而在更薄毛利下推高营收。
这一点尤为现实，因为新一代AI加速器正快速进入数据中心，承诺到年底带来最高10倍的token性能效率提升。Nvidia即将推出的Vera Rubin平台预计再带来10倍的token效率增益，进一步增强推理服务器的盈利潜力。但即便如此，AI公司能否产生足够营收以覆盖其巨额投资，仍不确定。
结论：AI行业正处关键节点，成本效率与可得性已变得与技术创新同等重要。面对来自中国开发者的激烈竞争，OpenAI、Google与Anthropic正在重构其定价模型与产品能力。
本文来源：https://www.whats-ai.com/ai-companies-cut-costs-as-token-prices-drop-and-models-become-accessible
若有版权问题，请来信告知：13951431913#139.com 或在右方提交评论。
The artificial intelligence industry is experiencing a profound shift, characterized by intense competition and aggressive cost-cutting strategies that are reshaping the landscape of AI model pricing. Traditionally, companies like OpenAI and Google have charged high premiums for their cutting-edge models, but the emergence of low-cost, high-performance Chinese models is forcing a re-evaluation of this business model. Giants such as OpenAI, Google, and Anthropic are now prioritizing price reductions for their entry-level products to compete with the likes of Kimi K3 and DeepSeek V4 Flash. This trend not only challenges the old paradigm of high-margin pricing but also suggests a future where accessibility and cost-efficiency are just as critical as model sophistication.
A New Era of Cost-Driven Competition
Over the past few years, the AI landscape has evolved significantly due to both technological advancements and global economic pressures. While American companies have long dominated the field with their investments in frontier research and hardware (such as Nvidia's GPUs), Chinese developers have leveraged their industrial scale and government support to deliver models at much more affordable prices without compromising on performance. The release of Kimi K3 and DeepSeek V4 Flash exemplifies this shift, pushing Western companies to rethink their pricing strategies and model offerings.
For instance, OpenAI recently dropped the price of its base model, ChatGPT 5.6 Luna, by 80%, and reduced the cost of its mid-range model, 5.6 Terra, by 20%. This is part of a broader trend where the focus is shifting from merely being the best to being the most accessible and affordable.
The Shift from Exclusivity to Accessibility
The competitive pressure is forcing AI developers to prioritize "availability" over "exclusivity." OpenAI's recent moves reflect this philosophy: while the top-tier models remain expensive, the more affordable options have become significantly cheaper. For example, GPT 5.6 Luna now costs $0.20 per million tokens for input and $1.20 for output, a drastic drop from its previous rates of $2.50 and $15. This pricing strategy allows OpenAI to better compete with Chinese alternatives like DeepSeek V4, which offers input costs below $0.50 and output costs around $0.87.
Meanwhile, the more capable GPT 5.6 Terra is now priced at $2.00 for input and $12.00 for output per million tokens, slightly undercutting K3's rates. However, OpenAI has also introduced faster, high-capacity options (like GPT 5.6 Sol Fast) at higher price points to cater to specific needs, though the overall strategy seems to be democratizing the use of AI for standard, non-time-sensitive tasks.
Financial Constraints and the Quest for Efficiency
These aggressive pricing strategies are happening at a time when many AI companies are facing significant financial challenges. OpenAI, in particular, is struggling with declining profit margins and falling revenues. The decision to lower prices is partly a business imperative to maintain their market share against cheaper competitors.
Furthermore, OpenAI's heavy reliance on external cloud computing infrastructure (as opposed to building their own) exacerbates their financial pressure. The company has committed hundreds of billions of dollars to cloud providers, which is necessary given the massive scale of data training and processing required. This financial burden underscores the critical need for these companies to improve the efficiency of their models to offset the high costs of operation.
The Jevons Paradox and the Future of AI
A key question arises regarding the sustainability of this low-cost model: will the savings in token costs actually lead to increased usage and profitability, or will it simply drive down the average revenue per user? This relates to the Jevons Paradox, which suggests that making things cheaper can sometimes increase consumption, potentially offsetting the loss in per-unit profit with increased volume.
The reality is that the industry is on the cusp of a significant efficiency boost. New generations of AI accelerators and processors promise to increase token performance by orders of magnitude within the next year. If these efficiency gains materialize, AI companies may be able to sustain low prices while maintaining healthy revenues. However, until then, the industry remains in a precarious state where the race to the bottom on price could have unintended consequences for the bottom line.
Conclusion: The Battle for Cost-Effective Supremacy
The AI industry is currently at a crossroads. The focus is shifting from pure technological superiority to a balance between capability and cost-effectiveness. The fierce competition from Chinese developers has acted as a catalyst, forcing established players to adapt their business models and find new ways to generate revenue without sacrificing performance. As the technology continues to improve, the future will likely belong to those who can offer the best value proposition, combining high-quality results with affordable prices.