SemiAnalysis says Claude plans give about 5x the API value of OpenAI's
SemiAnalysis says OpenAI halved the tokens on its $200 plan and cut cached pricing, leaving Anthropic's Opus 5.5 subscriptions with roughly five times the API-equivalent value of GPT 6.1 Sol.
Anthropic's subscriptions deliver about five times more API-equivalent value per month than OpenAI's for agentic work, comparing Claude Opus 5.5 with GPT 6.1 Sol, according to an analysis SemiAnalysis published on 5 October. The firm is an independent research shop, and the headline figure is its own estimate. Neither OpenAI nor Anthropic had commented in the coverage read.
The firm says OpenAI recently halved the number of tokens included on its $200 plan for each model tier and cut cached-input pricing for GPT 6.1 Sol, which together lowered API-equivalent value by more than 50 percent. A new $500 plan gives only 21 percent more of the Astra model than the old $200 one, it says. Existing $200 subscribers keep their old limits until 29 October.
To measure it, SemiAnalysis says it ran controlled experiments to see how each plan's usage meter moves for input, cache-write, cache-read and output tokens. It converted those rates into API dollars using its own September workload. The article says Anthropic's value per dollar stays consistent across tiers, while OpenAI's tiers are now similar to one another.
It also says Plans from the Chinese labs MiniMax, Moonshot and Z.ai beat OpenAI's new pricing but trail Anthropic's slightly, and that Cursor and Cognition plans offer worse value than first-party ones. Details beyond that sit behind the paywall, and the figures rest on one firm's workload. Different usage patterns could give different ratios, and no one else has replicated them.