不可持续的补贴
The Unsustainable Subsidy
三大AI厂商定价策略出现显著分化：Google的AI价格年增两倍，但其旗舰模型Gemini 3.1 Pro（输入2美元、输出12美元）仍是市场最低价；OpenAI的GPT-5.5（5美元/30美元）经历短期补贴后再次涨价；Anthropic的Claude Opus 4.7（5美元/25美元）价格保持稳定，且对最强模型有所下调。这些变动标志着行业正从不惜成本的市场份额争夺，转向注重利润与现金流管理的商业理性，以应对持续高昂的资本支出压力。
Tunguz 把三家模型厂的定价画成曲线，涨价的真相一目了然——烧钱抢份额的阶段结束了，做 AI 产品的都该重新算算成本模型。
In short : Three years of LLM pricing data: flagship prices dropped, then flipped upward, while budget models bounced back. Tokenizer efficiency softened the headline, but real costs are rising.
Google’s AI triples in price each year.
OpenAI’s flagship model was seemingly subsidized for a while, before rising again.
Anthropic’s AI has been the same price for a little bit & decreased for the most powerful models.
Those are three very different pricing strategies. If we compare the absolutes, the data completes the picture.
| Vendor | Model | Input ($/1M) | Output ($/1M) |
|---|---|---|---|
| Gemini 3.1 Pro | $2.00 | $12.00 | |
| Anthropic | Claude Opus 4.7 | $5.00 | $25.00 |
| OpenAI | GPT-5.5 | $5.00 | $30.00 |
Google remains the low-cost player, increasing the price on all its models but still less than half of the competition. Anthropic had maintained a luxe pricing until late last year.
The pricing changes indicate changes in strategy : cuts when cash is plentiful & share matters. Increases when cash is tight & margins matter. The latter is the case for all three vendors now when capex spending continues to set records.
来源：Tomer Tunguz 博客（VC 分析） · tomtunguz.com