Industry research firm SemiAnalysis compared subscription plans from leading AI labs and concluded that Anthropic is "an overwhelmingly better deal," offering about five times the API-equivalent value across mid-tier models like Claude Opus 5.5 and OpenAI's GPT-6.1 Sol. The analysis follows changes to OpenAI's $200-per-month plan that SemiAnalysis says effectively halved its value. At the $20 tier, Claude Pro provides 2.9 billion tokens per month, versus 1 billion for ChatGPT.
However, the picture shifts when measuring cost per task. Artificial Analysis tracks the estimated dollar cost of completing a specific task, and by that yardstick GPT-6.1 Sol ($0.72) is far more affordable than Claude Opus 5.5 ($5.98) at API rates. SemiAnalysis acknowledges that raw token volume may be more appropriate in some cases, and benchmark performance may not reflect real workloads.
Subscriptions are heavily subsidized, with Anthropic's plans consuming about 42 percent of its inference compute while generating only 10 percent of revenue. That subsidy is part of a marketing funnel, according to Skydive co-founder Dhruv Amin, but it doesn't scale for enterprises. With 93 percent of businesses exceeding AI budgets, Amin says many companies are shifting workloads to open source models to cut costs, sometimes by 75 percent or more.