Cursor Basics
Claude Opus 5 in Cursor: Pricing, Context and When to Use It
Claude Opus 5 is Anthropic's latest Opus model in Cursor and replaces Opus 4.8. Cursor rates it on par with Fable 5 on CursorBench while charging Opus rates: $5 per million input tokens and $25 per million output. It draws from the third-party Other Models pool, carries a 300k context window (1M max), and unlike Fable 5 it works under Zero Data Retention.
On this page
What is Claude Opus 5 in Cursor?
Claude Opus 5 is Anthropic's latest Opus model and the direct replacement for Opus 4.8 in Cursor's model picker. Cursor describes it as a step-change over Opus 4.8 on coding with the agent, professional knowledge work and long-horizon reasoning, and says it lands on par with Fable 5 on CursorBenchCursor's internal benchmark that scores models on both task performance and token efficiency, not accuracy alone. Press Enter for the full definition. while charging Opus pricing. Cursor's own recommendation is to run the high thinking variant for the best results.
The model card is the fastest way to see where it sits against the rest of the picker, so start there before the pricing detail.
- Model ID
- claude-opus-5 (Fast tier: claude-opus-5-fast)
- Provider
- Anthropic — a third-party model, not one of Cursor's first-party models
- Context window
- 300k, with a 1M maximum
- Capabilities
- Agent and Thinking
- Cursor's tiering
- Speed: medium. Cost: high. Intelligence: frontier.
- Usage pool
- The third-party Other Models pool
Per the model card at cursor.com/docs/models/claude-opus-5, checked 2026-07-27.
This is covered hands-on in Cursor First Hour — 4 short modules, free to read.
What is Claude Opus 5 good and bad at?
Cursor publishes both sides for this model, which makes the selection decision unusually concrete. The strengths cluster around sustained, multi-step work rather than raw single-shot answers, and the stated limitations are about cost and verbosity rather than capability.
- Cursor's claim
- Stronger than Opus 4.8 on thoroughness, carefulness and intent understanding
- What it means when you are working
- It holds the goal across long sessions and finishes multi-step work with less hand-holding
- Cursor's claim
- Competitive with Fable 5 on hard coding work
- What it means when you are working
- It often finishes tasks faster end-to-end while charging standard Opus rates
- Cursor's claim
- Strong at planning and tool use
- What it means when you are working
- It maps work before executing, chains tool results into follow-up actions, and adapts when tool output surprises it
- Cursor's claim
- Same per-token price as Opus 4.8
- What it means when you are working
- It still drains the Other Models pool faster than Sonnet 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.
- Cursor's claim
- Can over-elaborate in long sessions
- What it means when you are working
- On tasks where brevity matters more than depth, expect to trim
| Cursor's claim | What it means when you are working |
|---|---|
| Stronger than Opus 4.8 on thoroughness, carefulness and intent understanding | It holds the goal across long sessions and finishes multi-step work with less hand-holding |
| Competitive with Fable 5 on hard coding work | It often finishes tasks faster end-to-end while charging standard Opus rates |
| Strong at planning and tool use | It maps work before executing, chains tool results into follow-up actions, and adapts when tool output surprises it |
| Same per-token price as Opus 4.8 | It still drains the Other Models pool faster than Sonnet 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. |
| Can over-elaborate in long sessions | On tasks where brevity matters more than depth, expect to trim |
Strengths and limitations as published by Cursor, checked 2026-07-27.
Read the second row as a scheduling argument rather than a pricing one. If Opus 5 and Fable 5 land in the same place on hard coding work but Opus 5 gets there sooner at Opus rates, the saving shows up in elapsed time and in the tokens a longer session would have burned, not in a lower headline rate.
Does Claude Opus 5 work with Zero Data Retention?
Yes, and Cursor calls this out as a difference from Fable 5 rather than as a general property of the Claude family. For teams whose model allow-list is set by a security review rather than by benchmark scores, this is usually the deciding line in the whole document.
Cursor states that Opus 5 is Zero Data Retention compatible and that, unlike Fable 5, it does not require an Anthropic data-retention opt-in.
If your organisation runs under ZDRZero Data Retention. A contractual guarantee that the model provider won't store your code or train on it. Press Enter for the full definition., that turns a two-model choice into a one-model choice regardless of how the benchmarks compare — and it is the reason an Opus 5 approval can clear review when a Fable 5 approval stalls.
How much does Claude Opus 5 cost in Cursor?
Opus 5 draws from the third-party Other Models pool, which is metered separately from Cursor's first-party pool. Individual plans include at least $20 of third-party Other Models usage each month, and more on higher tiers. All rates below are per million tokens.
- Tier
- Claude Opus 5
- Input
- $5
- Cache write
- $6.25
- Cache read
- $0.5
- Output
- $25
- Tier
- Claude Opus 5 (Fast mode)
- Input
- $10
- Cache write
- $12.5
- Cache read
- $1
- Output
- $50
| Tier | Input | Cache write | Cache read | Output |
|---|---|---|---|---|
| Claude Opus 5 | $5 | $6.25 | $0.5 | $25 |
| Claude Opus 5 (Fast mode) | $10 | $12.5 | $1 | $50 |
Published rates at cursor.com/docs/models/claude-opus-5, checked 2026-07-27. Verify before budgeting.
The Fast tier (claude-opus-5-fast) was available at launch and exists for higher-priority output, at double the standard input and output rates. On legacy request-based plans it requires Max Mode. Cursor's guidance is to use it selectively, for time-sensitive or critical work rather than as a default.
Does a large prompt cost more with Claude Opus 5?
No, and this is the detail most worth knowing if you have budgeted for other models in the picker. Several models change their rates once a prompt gets large enough; Opus 5 does not.
Cursor states that all Opus 5 prompts bill at the base per-token rates, including when context goes above 300k, and that context windows up to 1M tokens use the same rates. There is no separate long-context multiplier for this model.
That is the opposite of the rule on GPT-5.6 Sol, where input above 272k tokens doubles the input rate and raises output to 1.5x. Check the rule per model rather than assuming your team's mental model carries across the picker.
Opus 5 also supports a thinking variant for deeper reasoning, and Cursor recommends the high setting for the strongest results — so the realistic cost question is how many thinking tokens a hard task consumes at $25 per million output, not whether a large repository dump triggers a penalty rate.
When should I pick Claude Opus 5 over another model?
Cursor's tiering puts Opus 5 at high cost and medium speed, which sets the boundary: it earns its rate on work that would otherwise need several attempts with a cheaper model. The pairings below follow directly from Cursor's published strengths and limitations rather than from a general preference.
- Situation
- Long multi-step work you want finished without hand-holding
- Reasonable pick
- Claude Opus 5, high thinking
- Why
- Cursor's stated strength: holding the goal across long sessions
- Situation
- Hard coding work where end-to-end time matters
- Reasonable pick
- Claude Opus 5
- Why
- Cursor rates it competitive with Fable 5, often faster, at standard Opus rates
- Situation
- The same work under Zero Data Retention
- Reasonable pick
- Claude Opus 5
- Why
- It does not require the Anthropic data-retention opt-in that Fable 5 does
- Situation
- High-throughput or routine edits
- Reasonable pick
- A Flash- 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.-class model
- Why
- Opus 5 drains the Other Models pool faster than Sonnet 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. for no accuracy you would notice
- Situation
- Short answers where brevity beats depth
- Reasonable pick
- A smaller model
- Why
- Cursor notes Opus 5 can over-elaborate in long sessions
| Situation | Reasonable pick | Why |
|---|---|---|
| Long multi-step work you want finished without hand-holding | Claude Opus 5, high thinking | Cursor's stated strength: holding the goal across long sessions |
| Hard coding work where end-to-end time matters | Claude Opus 5 | Cursor rates it competitive with Fable 5, often faster, at standard Opus rates |
| The same work under Zero Data Retention | Claude Opus 5 | It does not require the Anthropic data-retention opt-in that Fable 5 does |
| High-throughput or routine edits | A Flash- 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.-class model | Opus 5 drains the Other Models pool faster than Sonnet 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. for no accuracy you would notice |
| Short answers where brevity beats depth | A smaller model | Cursor notes Opus 5 can over-elaborate in long sessions |
Mapped from Cursor's published strengths, limitations and cost tiering.
On Teams and Enterprise plans there is one more option worth weighing: instead of pinning Opus 5 for everyone, Cursor Router can send only the hard requests to a model at this tier and keep routine work off it. That is the same argument the router's launch data makes, measured against pinning everything to Opus 4.8.
Frequently asked questions
What is Claude Opus 5 in Cursor?
Claude Opus 5 is Anthropic's latest Opus model and replaces Opus 4.8 in Cursor. Cursor describes it as a step-change over Opus 4.8 on coding with the agent, professional knowledge work and long-horizon reasoning, and rates it on par with Fable 5 on CursorBench at Opus pricing. It has a 300k context window with a 1M maximum, supports Agent and Thinking, and Cursor recommends the high thinking variant.
How much does Claude Opus 5 cost in Cursor?
Per million tokens, Cursor lists Opus 5 at $5 input, $6.25 cache write, $0.5 cache read and $25 output. The Fast tier (claude-opus-5-fast) doubles those to $10 input, $12.5 cache write, $1 cache read and $50 output. It draws from the third-party Other Models pool, and individual plans include at least $20 of that usage each month.
Does Claude Opus 5 charge extra for long context?
No. Cursor states that all Opus 5 prompts bill at the base per-token rates, including when context goes above 300k, and that context windows up to 1M tokens use the same rates. There is no separate long-context multiplier. This differs from GPT-5.6 Sol, where input above 272k tokens doubles the input rate and raises output to 1.5x.
Is Claude Opus 5 compatible with Zero Data Retention?
Yes. Cursor states Opus 5 is Zero Data Retention compatible and that, unlike Fable 5, it does not require an Anthropic data-retention opt-in. For teams whose model allow-list is set by a security review, that is often the deciding difference between the two.
When should I use Claude Opus 5 instead of a cheaper model?
Use it for long, multi-step work you want finished with less hand-holding, and for hard coding tasks where finishing sooner matters. Cursor notes the tradeoffs directly: it costs the same per token as Opus 4.8 and drains the Other Models pool faster than Sonnet or Composer, and it can over-elaborate in long sessions where brevity matters more than depth.
Sources & last verified
Cursor ships frequently. Facts verified against primary sources on July 27, 2026.
Keep reading
Rather do it than read about it? Run 11 interactive Cursor walkthroughs in a simulated editor. Free, no account needed.