Anthropic's best AI model struggles to attract users as cheaper tools thrive
Source: https://simonwillison.net/2026/Aug/23/anthropics-best-ai-model-struggles-to-attract-users-as-cheaper-t/
Author: Simon Willison · Published: 2026-08-23
Fetched: 2026-08-26T15:43:01Z via opencli web read
English Summary
Simon Willison merges the FT report with the Ramp AI Index (billing data from 70,000 card-spending companies): Anthropic's annualized revenue reached $65bn in July (up from $47bn in May) with 6,000 customers spending $100k+/yr and Q3 profitability expected; OpenAI's annualized revenue jumped 35% in the quarter to over $40bn after GPT 5.6's July launch. But Ramp's July spend breakdown flips the capability story: Opus 4.8 leads at 28.0% while the July-24-released Opus 5 holds only 3.5% and Fable 5 8.0% — price, not capability, sets the default model.
中文摘要
Simon Willison 把 FT 报道与 Ramp AI Index（7 万家企业的刷卡账单数据）合并成一张模型市场视图：Anthropic 7 月年化营收涨到 $65bn（5 月还是 $47bn），预期 Q3 盈利，有 6000 家客户年付 ≥$10w；OpenAI 年化营收季内涨 35% 至 $40bn+，GPT 5.6 七月发布后明显反弹。但 Ramp 的 7 月账单数据翻转了叙事：Opus 4.8 以 28% 占据第一，7 月 24 日才发布的 Opus 5 只占 3.5%，Fable 5 占 8%——单价把「最强模型」挤出了主力位。与已收录的 FT 原文条目（75a73f0d）互补，这里的增量是完整的模型份额明细表：价格梯度决定默认模型。
Why It Matters
模型选型与 API 成本优化的直接实证：企业真金白银的账单结构显示「足够好 + 便宜」挤出「最强」，做定价与选型决策的人应对照自查。
Obsidian Evidence
- AK-RSS Digest 摘要（OpenClaw定时任务/AK-RSS-Digest（89源精选）/2026-08-26-AK-RSS-Digest.md）第 4 条原文摘录：「Opus 4.8 占 28% 排第一，新发布的 Opus 5 只占 3.5%，Fable 5 占 8%」。
Provenance
- 选题来源: OpenClaw定时任务/AK-RSS-Digest（89源精选）/2026-08-26-AK-RSS-Digest.md
- 正文由 opencli web read 抓取（存档于 AAIF 工作目录），本页摘要与判断锚定在抓取正文与 digest 摘要上。
- 本页由 AAIF daily-intake-evening 任务生成（2026-08-26）。