FT Report: Anthropic's Priciest Flagship Model Stalls at 11% of Enterprise Spend

Citing Ramp spending data covering 70,000 firms, the FT reports that Anthropic's biggest, priciest flagship model (referred to as "Fable 5" in the article) has stalled at roughly 11% of enterprise AI spend and only 6% of token purchases more than two months after launch, as cheaper models prove sufficient for most tasks. Gary Marcus shared the report to mock the market's $2 trillion AI valuation expectations.

Confirmed

• Per multiple posts (@kimmonismus, @firstadopter, @danielmac8, @AccBalanced, @BurggrabenH), the FT report says the flagship holds only 11% of enterprise AI spend on Ramp, with growth stalled.
• @rohanpaulai and @FinanceYF5 add: Fable 5 accounts for only 6% of token purchases and costs about twice as much as Opus 5; enterprises match models to task difficulty. He also cautions that low flagship demand doesn't necessarily mean poor economics, since compute costs for smaller models fall faster.
• @firstadopter notes OpenAI's annualized revenue grew 35% this quarter, regaining share from Anthropic.
• @eyishazyer and @量子位 cite Ramp's AI Index: in July Fable 5 accounted for 6% of tokens and 11.4% of spend, while "Sol" took 25% of tokens and 23% of spend; Anthropic's enterprise revenue is a fraction of OpenAI's.
• @codestar and @zainhas note Opus 4.8 usage still exceeds Opus 5; @codestar attributes this to enterprise inertia (models locked into CI/CD pipelines), and @HanchungLee describes the same "model lock-in" dynamic.
• @zainhas argues Anthropic is testing the price elasticity of frontier intelligence and appears to have hit a ceiling; @Borthwick and @danielmac8 relay the report's attribution to high pricing and users accepting "good enough" performance.

Unconfirmed

• Model names like "Fable 5" and "Sol" appear to be FT pseudonyms, likely referring to Anthropic and OpenAI flagships; version numbers such as "Opus 4.8"/"Opus 5" lack corroboration.
• @eyishazyer mentions economists questioning Ramp data reliability, without elaboration.

Why it matters

• Enterprise payment data is a proxy for real willingness to pay: weak demand for premium flagships, squeezed by cheaper or open-source alternatives, pressures the high-valuation narrative across the industry—suggesting the AI race may be shifting from "who is smarter" to "who is cheaper." Some note model lock-in also suppresses flagship adoption.

2026-08-23 ~ 2026-08-24 · 19 related posts

• Anthropic's top model accounts for just 11% of corporate AI spend on Ramp — kimmonismus · 2026-08-23
• FT: Anthropic's top model accounts for just 11% of enterprise spending — daniel_mac8 · 2026-08-23
• [source] Ramp data shows low Claude share amid "high" mode downgrade controversy — eyishazyer · 2026-08-23
• Gary Marcus: Anthropic's Best Model Struggles to Attract Users — GaryMarcus · 2026-08-23
• [source] Anthropic's Priciest Model Sees Just 11% Adoption — BurggrabenH · 2026-08-23
• Data Suggests OpenAI Is Regaining Market Share from Anthropic — firstadopter · 2026-08-23
• [source] FT: Anthropic's Flagship Model Stuck at 11% of Enterprise AI Spend — firstadopter · 2026-08-23
• Anthropic struggles to attract users as cheaper AI tools thrive — Borthwick · 2026-08-23
• FT: Enterprises shift away from most powerful models, threatening frontier lab business models — AccBalanced · 2026-08-24
• Enterprise Inertia Keeps Opus 4.8 Ahead of Opus 5 in Usage — code_star · 2026-08-24
• Anthropic's Flagship Model Accounts for Only 11% of Enterprise Spending, Cheaper Models Win — AccBalanced · 2026-08-24
• Anthropic hits price elasticity limit: Fable is just 11% of enterprise sales — zainhas · 2026-08-24
• Fable 5 Makes Up Only 6% of Anthropic Sales — rohanpaul_ai · 2026-08-24
• Anthropic's Fable5 Struggles: Sales Just 11%, Growth Lags Behind OpenAI — 量子位 · 2026-08-24
• Enterprise inference inertia: why older models stay deployed — HanchungLee · 2026-08-24
• Anthropic's Fable 5 Market Share Slips to Cheaper Open-Source Models — inductionheads · 2026-08-24
• Anthropic's Top Model Fable 5 Accounts for Only 6% of Customer Token Usage — FinanceYF5 · 2026-08-24

2 near-duplicate retellings: iamaliveix · PMinervini