Fable 5 adoption data is an early test of frontier AI pricing

Fable 5 adoption was limited in the model’s first month, according to figures reported by The Decoder from Ramp’s spending data. But the result does not demonstrate a broad ceiling on what businesses will pay for frontier AI: Anthropic suspended access to Fable 5 days after launch, and later introduced a lower-priced model it said approached Fable-level capability.

The Decoder reported that Fable 5 made up about 6% of tokens bought from Anthropic and 11.4% of spending on Anthropic models during its first month. It put pricing at roughly $10 per million input tokens and $50 per million output tokens, about twice the cost of other flagship models cited in its comparison.

Ramp economist Ara Kharazian attributed the weak early uptake to price, arguing that the additional performance did not justify the premium for many companies. That is an interpretation of one spending snapshot, not a controlled measure of price sensitivity or return on investment. The data, as described by The Decoder, came from Ramp’s proprietary token-spend-management product and tilted somewhat toward technology companies.

What does Fable 5 adoption say about frontier AI pricing?

The narrow conclusion is that a higher-priced model may struggle when buyers cannot connect its added capability to measurable value in their everyday work. The available data cannot separate that explanation from others, including limited use-case fit, disrupted access and product-level constraints.

Fable 5’s rollout was not a normal launch. Anthropic said it released Fable 5 on June 9, then suspended access for all users after U.S. export controls took effect on June 12. The company said the controls were lifted June 30 and global access returned July 1. Anthropic also said it changed its safety systems after a reported safeguard bypass, acknowledging that the new classifier could more often flag benign coding and debugging requests.

That sequence makes a first-month usage reading difficult to treat as a clean measure of sustained enterprise demand. The model was unavailable for much of the period, while its post-restoration behavior included added safeguards that Anthropic said could create false positives in some routine work.

Cheaper capability changes the procurement comparison

Anthropic added another complication on July 24 with Claude Opus 5. The company said Opus 5 came close to Fable 5’s frontier intelligence at half the price, and claimed performance within 0.5% of Fable 5’s peak score on one coding evaluation at half the cost per task. Those are Anthropic product claims, rather than independent evidence of customer value, but they sharpen the buying decision: enterprises can compare premium access with a lower-cost option inside the same vendor’s lineup.

Broader payment-based measures still show business AI use expanding. In an April update, Ramp said 50.4% of businesses in its index paid for AI in March, up from 35% a year earlier. That figure predates Fable 5’s June release and measures overall paid AI use, so it cannot establish demand for the model. It does, however, argue against treating one disrupted launch as a verdict on the wider market.

For model vendors, the relevant test is whether an expensive system can show value beyond benchmark results in a specific workflow. Fable 5’s early data suggests that proof was not yet clear in Ramp’s observed sample.

This story draws on original reporting from The Decoder.