Anthropic's best AI model struggles to attract users as cheaper tools thrive Anthropic最强AI模型难以吸引用户，更便宜的工具却大受欢迎

TL;DR

• Anthropic's flagship Opus 5 model shows surprisingly low adoption (3.5% of model spend) despite being the company's best offering, while cheaper models like Opus 4.8 dominate at 28.0%
• Anthropic's annualized revenue surged to $65bn in July 2026 (up from $47bn in May), with 6,000 enterprise customers spending $100,000+ annually
• OpenAI's annualized revenue reached over $40bn after a 35% quarterly jump, boosted by the GPT-5.6 launch in July
• The Ramp AI index, derived from billing data across 70,000 companies, provides a novel real-world metric for tracking AI model adoption patterns
• Cost-performance tradeoffs are driving enterprise users toward older, cheaper models rather than the latest flagship releases

Why It Matters

This data reveals a critical market dynamic: the most capable AI models are not necessarily the most adopted, challenging the assumption that capability alone drives enterprise purchasing decisions. For AI practitioners and investors, it underscores the importance of pricing strategy and the growing maturity of the AI market, where cost efficiency increasingly competes with raw capability as a deciding factor.

Technical Details

• Ramp AI Index: A novel adoption tracking methodology using billing data from 70,000 Ramp credit card-using companies to estimate real-world model usage patterns, offering a more grounded alternative to self-reported metrics
• Anthropic Model Spend Breakdown (July 2026): Opus 4.8 leads at 28.0%, followed by Sonnet 4.6 (8.3%) and Fable 5 (8.0%), while the newly released Opus 5 (released July 24th) captured only 3.5% — suggesting limited enterprise uptake in its first week
• Revenue Trajectory: Anthropic's annualized revenue grew from $47bn (May) to $65bn (July), while OpenAI's crossed $40bn with 35% quarterly growth, indicating intense competitive pressure
• Enterprise Concentration: 6,000 Anthropic customers spending $100,000+ annually represents a highly concentrated revenue base, raising questions about customer diversification risk

Industry Insight

• Price sensitivity is reshaping model adoption: Enterprises are prioritizing cost-effective solutions over cutting-edge capability, suggesting that model pricing strategies will be as critical as performance benchmarks in determining market success
• The "flagship paradox": Releasing a more expensive, marginally better model risks cannibalizing adoption of proven, cheaper alternatives — companies should consider tiered pricing or gradual capability rollouts to avoid alienating cost-conscious customers
• Alternative data sources like the Ramp AI index will become essential: As self-reported metrics lose credibility, billing-based adoption tracking offers a more reliable signal for investors and strategists evaluating AI company performance

摘要

Anthropic 的旗舰模型 Opus 5 尽管是公司最优质的产品，但采用率却出人意料地低（仅占模型支出的 3.5%），而 Opus 4.8 等更便宜的模型则占据主导地位，占比达 28.0%
Anthropic 2026 年 7 月的年化收入飙升至 650 亿美元（5 月为 470 亿美元），拥有 6,000 家年支出超过 10 万美元的企业客户
OpenAI 年化收入在 7 月推出 GPT-5.6 的推动下，季度环比增长 35%，突破 400 亿美元
Ramp AI 指数源自 70,000 家公司的账单数据，为追踪 AI 模型采用模式提供了一种新颖的现实指标
成本与性能的权衡正推动企业用户转向更老旧、更便宜的模型，而非最新的旗舰产品

深度分析

核心要点

• Anthropic 的旗舰模型 Opus 5 尽管是公司最优质的产品，但采用率却出人意料地低（仅占模型支出的 3.5%），而 Opus 4.8 等更便宜的模型则占据主导地位，占比达 28.0%
• Anthropic 2026 年 7 月的年化收入飙升至 650 亿美元（5 月为 470 亿美元），拥有 6,000 家年支出超过 10 万美元的企业客户
• OpenAI 年化收入在 7 月推出 GPT-5.6 的推动下，季度环比增长 35%，突破 400 亿美元
• Ramp AI 指数源自 70,000 家公司的账单数据，为追踪 AI 模型采用模式提供了一种新颖的现实指标
• 成本与性能的权衡正推动企业用户转向更老旧、更便宜的模型，而非最新的旗舰产品

重要意义

这一数据揭示了一个关键的市场动态：最强大的 AI 模型并不一定是采用率最高的模型，这挑战了"能力 alone 驱动企业采购决策"的假设。对于 AI 从业者和投资者而言，这凸显了定价策略的重要性，以及 AI 市场日益成熟——成本效率正与原始能力竞争，成为决定因素。

技术细节

• Ramp AI 指数：一种新颖的采用率追踪方法，利用 70,000 家使用 Ramp 信用卡公司的账单数据来估算现实世界中的模型使用模式，提供了比自我报告指标更接地气的替代方案
• Anthropic 模型支出细分（2026 年 7 月）：Opus 4.8 以 28.0% 领先，其次是 Sonnet 4.6（8.3%）和 Fable 5（8.0%），而新发布的 Opus 5（7 月 24 日发布）仅占 3.5%——表明其在首周的企业采用有限
• 收入轨迹：Anthropic 年化收入从 5 月的 470 亿美元增长至 650 亿美元