The numbers arrived quietly, tucked inside a Ramp report that most of the industry skimmed and discarded. Fable 5 — the model Anthropic billed as its most powerful creation — accounts for just 11.4% of the company's enterprise spending and a mere 6% of its token usage. The cheaper Opus 5, launched weeks later, has already overtaken it in adoption. Code is law, but people are the protocol. And the people with procurement budgets have just voted. They rejected the most sophisticated model on the market, not because it failed on benchmarks, but because its marginal intelligence gain did not justify a 2x price multiplier. As someone who spent the 2022 Bear Market watching crypto teams learn similar lessons about value versus hype, this moment carries a familiar weight.

The Two-Tier Strategy That Became a Two-Tier Reality

Let me frame this with the precision this story deserves. Anthropic executed what I would call a dual-flag strategy. Fable 5, priced at $10 per million input tokens and $50 per million output tokens, stands at double the cost of Opus 5. The company's own positioning materials describe Opus 5 as “near-Fable 5 frontier intelligence at half the price." That single sentence is the quietest admission in the industry: the gap between the flagship and the sub-flagship is not a generation gap. It is a half-step. This aligns with the broader law of diminishing returns we have documented since DeFi Summer — moving from 80 to 90 in benchmark scores is expensive; moving from 90 to 95 is proportionally far more expensive. And enterprises, it turns out, understand this arithmetic better than the marketing departments.

Ramp's data gives us a working elasticity estimate. A 100% price premium yielded roughly an 87% reduction in spending share. That is a demand elasticity of around 4.5 — the highest we have seen in enterprise software. The token-spend divergence is even more telling. Fable 5 accounts for 6% of token usage but 11.4% of spend, meaning the customers who do use it are paying the premium. Yet the base is too small to generate network effects. In my work with governance systems, this looks familiar: a small group of high-conviction participants willing to pay for excellence, surrounded by a silent majority that wants “good enough.”

The Contrarian Reading: Why the “Weakest” Model May Be Anthropic's Strongest Asset

Here is where the conventional analysis gets it wrong. Most coverage frames Fable 5's low adoption as a failure. I read it as the first clear evidence that AI has entered its value-differentiated phase. This is exactly the pattern we saw with Ethereum's Layer 2 adoption curve — the rollup with the most advanced cryptography does not always win; the one with the cheapest data availability does. 99% of rollups do not generate enough data to justify a dedicated DA layer. Similarly, 90% of enterprise workloads do not require Fable 5's frontier intelligence. They need reliability, consistency, and cost predictability. Opus 5 delivers that. The market is not rejecting Anthropic. It is rejecting the notion that intelligence alone deserves a premium.

Accel partner Miles Clements articulated what enterprises already believe: “Most people do not need frontier models on an ongoing basis.” That statement will be the epitaph of the AI supremacy race. GPT-5.6 Sol has captured 25% of OpenAI's enterprise token volume, which tells us OpenAI's pricing aligns better with perceived value. The architecture is simple: enterprises will pick the highest intelligence per dollar, not the highest intelligence. Governance isn't just about who votes; it is about what they choose with their resources. Here, the votes are clear.

A Personal Correction: The Lesson of the 2024 ETF Campaign

During the 2024 ETF transparency advocacy campaign, I worked with ten Asian universities to teach institutional crypto adoption. We kept telling students that regulation enhances decentralization. A more precise lesson emerged: institutional adoption is not about convincing funds to accept a token. It is about convincing the token to meet institutional cost-benefit thresholds. The same applies to AI models. Anthropic's Fable 5 is the equivalent of a sophisticated treasury protocol. Powerful, secure, but requiring significant capital for marginal utility. Opus 5 is the stablecoin — the asset that gets used daily because it is practical.

We didn't need a survey to understand the AI adoption paradox. We needed a payment processor to track how enterprises spend their compute budgets. Ramp provided that. And the data points to a structural shift: AI procurement is moving from technical specs to financial ROI. The adoption curve of Opus 5 in months, versus the slow start of Fable 5, is a natural experiment that mirrors what we observed in the 2022 bear market. When the tide went out, the projects that survived were not the ones with the most ambitious roadmaps. They were the ones that could prove unit economics.

The Hidden Risk: Governance and Security Decay

The more subtle risk is what this means for AI safety. Anthropic's brand is built on responsible AI development. Fable 5 presumably includes their most advanced alignment and red-teaming capabilities. When enterprises decline to deploy it, those safety capabilities are not being exercised at scale. Over time, this creates a public verification gap. The safest model is the one being used in production, not the one in the lab. In blockchain terms, this is the difference between a audited smart contract that holds no assets and one with a verified exploit history. The latter is more trustworthy.

The market is also witnessing a split between model capability and model monetization. That will impact how AI companies allocate compute resources. If Fable 5 only uses 6% of tokens, the inference compute is underutilized. Anthropic will have to reallocate those resources to Opus 5 — or eat the sunk cost. This is the classic capital reallocation problem we see in decentralized networks when an early fork loses adoption. The lesson: don't stay attached to the technical leader. Invest in the one that the community is actually using.

A Forward-Looking Takeaway

The Fable 5 story is not about a model failing. It is about the industry waking up to the notion that intelligence has no intrinsic value — only applied value. The next 12 months will determine whether Anthropic becomes the leader of the value segment or gets squeezed into the premium corner. If I were advising the company, I would not lower Fable 5's price. I would reposition it as the specialist tool for complex, high-stakes reasoning, while letting Opus 5 carry the revenue weight. The road ahead demands that we stop valuing the most powerful model and start building the most useful infrastructure. Governance isn't about choosing the most powerful validator; it is about choosing the one that keeps the network alive. The same logic applies to the models we build. What matters is not who has the best model in the lab, but who has the most-used protocol in production.