Tomasz Tunguz：AI 模型商品化后，实验室该如何调整策略
A Change in AI Strategy • Tomasz Tunguz
Tomasz Tunguz 认为 AI 模型已进入商品化阶段：自 7 月以来 token 用量增长 50%，价格下降 41%，前沿模型 token 份额从 8 月的 53% 降至 40% 多。他给出的应对策略是合作转售模型、控制用户界面以建立分发壁垒，并收集路由数据用于后续模型训练。文中提到 OpenAI 将 Luna 价格下调逾 80%，并与 Baseten 合作转售开放权重模型。
正文
当前显示原文，中文全文译文尚未就绪。
In short : AI models have reached commodity status as prices fall 41% while token usage surges 50%. To win relative share, labs should partner & resell models, control the user interface where distribution becomes the moat, & capture routing data to reduce the cost of training future models.
What if AI models have just reached commodity status? In early 2025, we debated whether labs would become the airlines of the AI industry : high cost, low margin businesses with undifferentiated products.
Over the past month, the evidence has mounted :
OpenAI cut prices on Luna by more than 80% to win share.1; Frontier models’ share of tokens slipped from 53% in August into the mid-40s as companies defaulted to smaller, cheaper tiers.2; Open source demand has shifted decisively from frontier-dominated to medium-&-small tiers.3; Spending among the top 1% of adopters fell 9.7% in August to $7,205 per employee per month, cooling from its July peak.4; Decision models like TypeSafe’s Jev are winning share of LLM calls at $0.04 per million input tokens, roughly 75x cheaper than standard LLMs for routing, gating, & classification.5
Throughout all this, token usage has surged 50% since July, while prices have fallen 41%.6
What’s the strategy in this environment?
First, partner. OpenAI announced a partnership with Baseten to resell open-weight models.7 Elon tweeted that the Grok bot would use the best model to complete a task, not just SpaceXSI models.8 Reselling produces a new, high-margin revenue stream by collecting a toll on products without bearing the cost to serve.
No company can provide the best model for every use case. Maximizing customer attention & retention is the most valuable asset in a commodity market.
Second, control the user interface. Distribution has become the moat. Build great harnesses that demand a lot of tokens : give them mononyms like Dots, Bot, & Muse to retain users.9 Also provide the best models for the job so people stay. This suggests ads & commerce as an important revenue driver for the B2C market.
Third, capture user data for intelligent routing & training infrastructure. Reselling other models while routing tasks across them aggregates a tremendous amount of information useful for subsequent model training.
Winning share becomes the only game that matters when aggregate dollar spend cools while token volume compounds. This means controlling the user interface.
OpenAI’s recent surge toward a ~$70b run rate highlights the dynamic : aggressive price cuts & marketplace aggregation allowed it to recapture share within a boat length of Anthropic.10
In a commoditizing market, the spoils do not accrue to the lab with a marginal benchmark lead, but to the platform that aggregates the volume.
GPT-5.6 & The Most Important Market in AI is the Middle ↩︎; Ramp September 2026 AI Index ↩︎; The Most Important Market in AI is the Middle & Segmentation Drives Market Share Wins in AI ↩︎; Ramp AI Index ↩︎; Jev by TypeSafe AI ↩︎; Ramp September 2026 AI Index ↩︎; Announcing our partnership with OpenAI ↩︎; Elon Musk on X ↩︎; ha! 😄 ↩︎; Segmentation Drives Market Share Wins in AI ↩︎