Claude Opus 5.5 in Cursor: Pricing and Uses

Claude Opus 5.5Anthropic's top Opus-tier model in Cursor at $4 per million input tokens and $20 per million output, 20% below Claude Opus 5, which it replaces. Press Enter for the full definition. is Anthropic's latest Opus model, and Cursor's docs list it as the replacement for Opus 5. Cursor says it sets a new high score on CursorBenchCursor's internal benchmark that scores models on both task performance and token efficiency, not accuracy alone. Press Enter for the full definition., ahead of Fable 5.1, while costing 20% less per token than Opus 5. The recommended setup is the high thinking variant. The model card below holds the rest of the identifying detail.

The word to hold onto in that card is replaces: the rates below are different from Opus 5's.

Cursor also gives Opus 5.5 the full set of agent tools: file search, web, file reads and edits, shell commands, the browser, image generation, clarifying questions and rule fetching. The docs list no tool that Opus 5.5 lacks.

Opus 5.5 is a third-party model, so it is not part of the usage included with Cursor plans. On The X SubscriptionThe subscription Cursor is sold through: one plan for Grok, Cursor and X, with Free, Plus, Premium, Super and Ultra tiers, per-seat Teams, a weekly included pool for Cursor models shared with Grok, and third-party models billed at API list price from prepaid credit. Press Enter for the full definition. it bills from your prepaid credit at the rates below, and on legacy Cursor plans it draws from the Other ModelsOn legacy Cursor plans, the monthly pool for supported third-party models from providers such as Anthropic, Google and OpenAI, separate from Cursor Models. On The X Subscription third-party models are billed from prepaid credit instead. Press Enter for the full definition. pool. All prices are per million tokens.

Against Opus 5 the drop is $5 to $4 on input and $25 to $20 on output, which is where the 20% comes from. The cache-read change is bigger in relative terms. Reads now cost $0.20, which is 0.05x the input rate where Opus 5 sat at 0.10x, so a read that was $0.50 is down 60%. I'd guess that is the line that shrinks a long session's bill the most, since the same context gets read back on every turn, though I can't tell you the split without your own usage numbers.

Fast mode (claude-opus-5-5-fast) was available at launch for higher-priority output and bills at $8 per million input and $40 per million output, exactly double the standard rates. On legacy request-based plans it requires Max Mode. Cursor's guidance is to use it selectively, for time-sensitive or critical work.

Global endpoints use the table rates. Regional and US-only endpoints are priced 10% higher, at $4.40 per million input tokens and $22 per million output tokens. Context is the other place a surcharge might be expected, and there isn't one: every Opus 5.5 prompt bills at the base per-token rates, including above 300k tokens, and windows up to 1M tokens use the same rates.

Cursor's models and pricing page adds that third-party models are available on Premium and above at API list price from the first request, paid from prepaid credit, and that Plus and Free do not include them. It also lists Opus 5.5 and its Fast mode as requiring Max Mode on legacy request-based plans, and notes that opting in to regional data residency adds a 10% uplift on model pricing for eligible models. The same table lists the Opus 5.5 cache write at $5 per million and the cache read at $0.20, matching the model card.

Cursor lists five strengths and one limitation for this model, which makes the choice unusually easy to reason about. The strengths below are Cursor's claims, not independent measurements, so read the table as what the vendor says to try rather than as a verdict.

The benchmark claim carries a qualifier worth reading twice: highest score at launch, at standard Opus pricing. That is a comparison with Fable 5.1, and the price half of the sentence matters as much as the score half. A top score at Fable pricing would be a different decision.

The communication point is easy to wave off. Cursor's wording is that the model explains what it did and why without over-elaborating, which it says makes long agent sessions easier to follow. That is Cursor's description, not a measurement.

Subagents are the third strength with a practical angle. Cursor says Opus 5.5 works well as the coordinator, delegating to subagents while keeping the parent task on track.

Yes. Cursor says Opus 5.5 has no data-retention requirements for general access, like earlier Opus models and unlike Fable 5.1. Teams that use Zero Data Retention with Opus today can keep that setup with Opus 5.5.

The docs frame it as a continuation: teams using Zero Data Retention with Opus today can keep that setup, and the contrast Cursor draws is with Fable 5.1, not with the Opus line.

For data residency, the same page points to Privacy and Data Governance. The Opus 5.5 page itself states that regional and US-only endpoints are priced 10% higher.

Cursor does not publish a threshold for when Opus 5.5 pays for itself, and I would not invent one. What the docs do give is a tier (cost high, speed medium) and one stated limitation, which is that it uses prepaid credit or the Other ModelsOn legacy Cursor plans, the monthly pool for supported third-party models from providers such as Anthropic, Google and OpenAI, separate from Cursor Models. On The X Subscription third-party models are billed from prepaid credit instead. Press Enter for the full definition. pool faster than Sonnet 5.5 or ComposerCursor's own fast coding model, tuned for the editor and priced well below frontier models; the recommended day-to-day model for executing a plan. Press Enter for the full definition.. The pairings below follow from those statements and go no further.

The last row reflects the one limitation Cursor lists: Opus 5.5 uses prepaid credit, or the Other ModelsOn legacy Cursor plans, the monthly pool for supported third-party models from providers such as Anthropic, Google and OpenAI, separate from Cursor Models. On The X Subscription third-party models are billed from prepaid credit instead. Press Enter for the full definition. pool on legacy plans, faster than Sonnet 5.5 or ComposerCursor's own fast coding model, tuned for the editor and priced well below frontier models; the recommended day-to-day model for executing a plan. Press Enter for the full definition..