Anthropic ships Claude Opus 5.5 at 40 percent lower cost than Opus 5

Sources: primary, benchmarks, discussion
Summary: Anthropic released Claude Opus 5.5 on 2026-09-22 as the first model in the Claude 5.5 family. Input and output tokens are priced at $4 and $20 per million, 20 percent below Opus 5, and cache reads at $0.20 per million, 60 percent below Opus 5. Anthropic states a 40 percent cost reduction on typical workloads at default settings, output generation more than 30 percent faster, and deployment under safeguards similar to Claude Fable 5.1 because capability in biology and cybersecurity is comparable.
Why it matters: Cache reads dominate token spend in agentic coding, so the $0.20 per million cache-read price moves the cost of long-running agent sessions more than the headline rate does.
Follow-up: Whether independent harnesses reproduce the vendor-run figures of Terminal-Bench 4.0 at 66.4 percent, FrontierCode v1.1 main at 54.4 percent, and CursorBench 4.0 at 57.8 percent, which Anthropic's own post frames as a less reliable guide at these capability levels. Also when Sonnet 5.5 and Haiku 5.5 ship.