Author(s): Kushal Banda

Originally published on Towards AI.

Anthropic’s new Opus costs 40% less than its predecessor, tops the benchmarks, and can’t turn its thinking off anymore. Good and bad, in one release.

At Quantium, a task ran 38 prompts over four days on Claude Opus 5. The same task, on Claude Opus 5.5, took 11 prompts in three hours. That’s not a benchmark number. It’s a CIO describing what changed on their own team, in the launch material Anthropic published on September 22, 2026.

After the introduction, the article explains what Opus 5.5 changes in practice—mostly efficiency and cost—while noting that “thinking” can no longer be disabled and that reasoning depth is controlled through an effort parameter. It covers updated pricing and cache costs, shows how Opus 5.5 is positioned against competing models on the Intelligence Index (including where it leads), and summarizes partner feedback emphasizing that users need “less from me” (fewer steps and tokens to reach the same outcomes). It also discusses potential downsides and tradeoffs: some criticism about Anthropic’s messaging around pacing the frontier, evidence that Opus 5.5 may use more tokens in at least one independent review despite the lower unit price, and lingering questions about whether communication-style improvements truly solve prior verbosity complaints. The piece concludes that for teams already running Opus 5, upgrading is likely straightforward on cost/output grounds, but the key operational “gotchas” are the breaking API/behavior changes related to thinking blocks and tool/use constraints.

Read the full blog for free on Medium.

Join thousands of data leaders on the AI newsletter. Join over 80,000 subscribers and keep up to date with the latest developments in AI. From research to projects and ideas. If you are building an AI startup, an AI-related product, or a service, we invite you to consider becoming a sponsor.

Published via Towards AI