Anthropic cuts Opus prices as Claude Opus 5.5 ships
Reviewed by Helen Jones
Anthropic released Claude Opus 5.5 on 22 September 2026 at $4 per million input tokens and $20 per million output tokens. That is 20 percent below Opus 5, which stayed at $5 and $25. Cache reads drop to $0.20 per million tokens from $0.50. The model has a 1 million token context window and up to 128,000 output tokens. Anthropic says typical default workloads cost about 40 percent less than Opus 5 because of lower rates and fewer tokens used, and that output is more than 30 percent faster. Fast mode is double the standard rates. Batch is $2 and $10. It is on the Claude API, Claude Code, Amazon Bedrock, Google Cloud Vertex AI, and Microsoft Foundry. Paid Pro, Max, Team, and Enterprise plans get it. Free does not. Claude Code 2.1.280 makes Opus 5.5 the default Opus model.
Until this week, Opus 5 held its list prices while long-context coding and agent sessions burned money mainly on cache reads, not on the headline input line. Founders who vibe-code or run prompt chains treated Opus as a premium default and paid for it even when most of the bill was rereading the same repo, docs, or conversation. The cut is the first listed Opus-generation price drop, and cache reads fell 60 percent, which is the part of the bill that actually dominates those loops. Fable 5.1 still sits at $10 and $50, so the expensive top tier did not move. The 40 percent typical-workload claim is Anthropic's own number and depends on token use, not list price alone.
Analysis
Treat this as a change to act on for cache-heavy Claude Code and long-context workflows, not a reason to route every sales-page draft through Opus. Run one real job you already pay for on Opus 5 and Opus 5.5, then keep 5.5 only if tokens, cache hits, wall time, and dollars all improve.
Source note
Pulse published by Collab365 Spaces, reviewed by Helen Jones on . Cite as "Anthropic cuts Opus prices as Claude Opus 5.5 ships", Collab365 Spaces. 3 sources referenced.