Gary Marcus 用数据告诉你，AI 推理价格一年内垮塌到原来的百分之一，中国模型成本只有 Anthropic 的零头，这对整个行业都是改变游戏规则的事。
Gary Marcus 指出 AI 推理成本正经历有史以来最陡峭的商品曲线。他引用数据，中国模型每百万 token 仅需 $0.50，比 Anthropic 的 $56 便宜 112 倍。相比之下，石油价格压缩用了 40 年，半导体用了 20 年，而 AI 推理仅用数月。Anthropic 和 OpenAI 的定价分别为 $56 和 $26，远高于 Meta 的 $1.50、xAI 和 Google 的 $1。
“the steepest commodity curve any technology has run in recorded history” - exactly what I have been...
“the steepest commodity curve any technology has run in recorded history” - exactly what I have been warning people about in Marcus on AI since August 2023. We had nearly three years to prepare for this moment, but almost nobody listened. Shruti @heyshrutimishra Chinese models are 112x cheaper than Anthropic per million tokens. Chamath laid it out on CNBC: a "barrel of intelligence" costs $56 from Anthropic, $26 from OpenAI, $1.50 from Meta, $1 from xAI and Google, and $0.50 from Chinese models. That is not a pricing quirk. That is the steepest commodity curve any technology has run in recorded history. Oil took 40 years to compress like this. Semiconductors took 20. AI inference is doing it in months. The companies sitting at $26 and $56 are not stupid. They're buying time… betting that trust, safety, and enterprise contracts hold the premium long enough for costs to catch up. What they cannot bet on is the timeline. Because the $0.50 model is not a demo. I've been inside the labs building it. Your browser does not support the video tag. 🔗 View on Twitter 🔗 View Quoted Tweet 💬 4 🔄 7 ❤️ 34 👀 3951 📊 8 ⚡