How Much Does Cursor Composer 2.5 Cost?
Cursor Composer 2.5 costs $0.50 per million input tokens and $2.50 per million output tokens on the Standard tier. The Fast tier, which Cursor ships as the default, costs $3.00 and $15.00 for the same work.
That is a six times difference on the same model name. Most people running Composer 2.5 have never opened the setting that controls it.
Composer 2.5 shipped on May 18, 2026 as Cursor's own in-house coding model rather than a rebadged Anthropic or OpenAI model. Cursor positions it as competing with Claude Opus 4.7 and GPT-5.5 at a fraction of the cost. On Cursor's published rates, that framing is strongest on Standard and weakest on the tier Cursor selects for you.
AI Perks tracks what AI coding tools actually cost and which vendors hand out credits to offset the bill.
Why the Default Fast Tier Costs Six Times the Standard Rate
Fast is a speed tier priced above Standard. Between Composer 2 in March and Composer 2.5 in May, that premium went from three times to six times, and the expensive side became the default.
Composer 2 priced Fast at $1.50 input and $7.50 output against a $0.50 and $2.50 Standard rate, a clean 3x. Composer 2.5 left Standard untouched and moved Fast to $3.00 and $15.00.
| Model and tier | Input per 1M | Output per 1M | Cache read per 1M | Multiple vs Standard |
|---|---|---|---|---|
| Composer 2.5 (Standard) | $0.50 | $2.50 | not confirmed | 1x |
| Composer 2.5 (Fast, default) | $3.00 | $15.00 | not confirmed | 6x |
| Composer 2 (Standard) | $0.50 | $2.50 | $0.20 | 1x |
| Composer 2 (Fast) | $1.50 | $7.50 | $0.35 | 3x |
Composer 2.5 figures come from Cursor's changelog entry of May 18, 2026. Composer 2 figures, including both cache read rates, come from its launch post of March 19, 2026. No cache read or cache write figure for Composer 2.5 is confirmed in either, so treat a 2.5 cache number quoted anywhere else as unverified until Cursor states it.
The practical takeaway is blunt. If latency is not your bottleneck, moving Composer 2.5 from Fast to Standard cuts the per-token rate by roughly 83%. That takes no plan change, no migration and no second vendor. It is a dropdown.
What Changed Between Composer 2 and Composer 2.5
Standard pricing did not move between the two releases. Fast doubled, and the default landed on it.
Composer 2 arrived on March 19, 2026 as Cursor's in-house frontier-class agentic coding model, positioned on benchmarks against Claude Opus 4.6. Composer 2.5 followed on May 18, 2026 as Cursor's most capable in-house coding model to date, positioned against Claude Opus 4.7 and GPT-5.5 at a fraction of their cost.
| What changed | Composer 2 (Mar 19, 2026) | Composer 2.5 (May 18, 2026) |
|---|---|---|
| Positioning | In-house frontier-class agentic coding model, pitched against Claude Opus 4.6 | Cursor's most capable in-house model to date, pitched against Claude Opus 4.7 and GPT-5.5 |
| Standard rate | $0.50 in / $2.50 out | $0.50 in / $2.50 out |
| Fast rate | $1.50 in / $7.50 out | $3.00 in / $15.00 out |
| Default tier | not stated at launch | Fast |
| Cache read | $0.20 Standard, $0.35 Fast | not confirmed |
The pattern worth budgeting around is that the cheap side of this price list held steady across two releases while the fast side did not. A Fast multiple that went from 3x to 6x inside two months is not a number to hardcode into a cost model.
What Cursor Has and Has Not Confirmed About Composer 2.5
The benchmark figures attached to Composer 2.5 are approximate, and several specs people quote confidently are not confirmed at all.
The two scores reported with the release put Composer 2.5 at roughly 79.8% on SWE-Bench Multilingual and 69.3% on Terminal-Bench 2.0, against Claude Opus 4.7 and GPT-5.5. Both are approximate figures reported with the launch, not precise numbers to quote to two decimal places in a procurement document.
What is not confirmed:
- Parameter count. A size of roughly one trillion parameters has been reported. Cursor has not confirmed it, so it stays a report rather than a spec.
- Context window. No context window figure for Composer 2.5 is confirmed in Cursor's Composer 2 or Composer 2.5 posts. Anyone quoting one is quoting something unverified.
- Cache pricing for 2.5. Confirmed cache read rates exist for Composer 2 only, at $0.20 and $0.35 per million.
- Composer 2's current status. Its March launch post does not settle whether that model is still selectable today or at what rate, so a pinned model in a config is worth verifying against your actual usage line.
That last one matters more than it sounds. A config written against March pricing carries March assumptions, and one of the two tiers in this article doubled in eight weeks.
What Each Cursor Plan Costs and What Composer Counts Against
Composer sits in Cursor's first-party models pool, which comes with usage included on the paid individual plans. The free Hobby plan exists separately, with limited usage.
| Plan | Price | Composer usage |
|---|---|---|
| Hobby | Free | Limited usage |
| Pro | $20/month | Included in the first-party models pool |
| Pro+ | $60/month | Included in the first-party models pool |
| Ultra | $200/month | Included in the first-party models pool |
| Teams | $40 per seat/month | Not detailed in Cursor's Composer posts |
Cursor publishes per-token Composer rates alongside that pool, which is why the tier setting matters more than the plan tier for most solo users. Going from Pro to Pro+ triples the subscription. Going from Fast to Standard cuts the token rate by about 83% and costs nothing.
Composer 2.5 also shipped with double usage for the first week, per Cursor's changelog. That window closed in May 2026. The levers that remain are tier selection and credits from the layer underneath, which is what AI Perks exists to map.
How to Cover the AI Coding Bill With Free Credits
Cursor does not publish a dedicated startup or Composer-specific credit program. The credits that meaningfully reduce an AI coding bill come from the model vendors and cloud providers underneath the editor.
- Tier selection first. Before any credit hunt, the Fast-to-Standard switch is the largest single reduction available on Composer spend, and it is available to every plan.
- Your own vendor keys. Startups can bring their own Anthropic or OpenAI keys funded by those vendors' startup programs. That moves frontier model spend off the Cursor pool and onto credits someone else granted.
- Quarterly rechecks. Cursor moved its Fast multiple from 3x to 6x across two releases in two months. A price assumption written in March was already wrong by May.
The programs worth mapping are the ones funding the API layer under any AI editor, rather than the editor subscription on top of it. See which ones apply to you at getaiperks.com.
Frequently Asked Questions
How much does Cursor Composer 2.5 cost per million tokens?
Composer 2.5 costs $0.50 input and $2.50 output per million tokens on Standard. The Fast tier, which is the default, costs $3.00 input and $15.00 output. Cursor has confirmed cache read rates for Composer 2 ($0.20 Standard, $0.35 Fast) but no cache read or cache write figure for Composer 2.5.
Is Composer 2.5 included in the Cursor Pro plan?
Yes. Composer sits in Cursor's first-party models pool, which comes with usage included on the paid individual plans: Pro at $20 per month, Pro+ at $60 and Ultra at $200. The free Hobby plan is separate and has limited usage. Plan comparisons at getaiperks.com.
Should I use Composer 2.5 Fast or Standard?
Standard, unless latency is your actual bottleneck. Fast is six times the per-token price, so switching cuts the rate by about 83%. Fast is the default, which means most Cursor users are paying the higher rate without having chosen it.
Is Composer 2.5 cheaper than the frontier models it is compared against?
Cursor positions Composer 2.5 as competing with Claude Opus 4.7 and GPT-5.5 at a fraction of their cost, and the Standard rate of $0.50/$2.50 is the basis for that claim. Any head-to-head number should be checked against what those vendors publish on the day you check, since model prices move. Cost tracking at getaiperks.com.
What happened to Cursor Composer 2?
Composer 2 launched on March 19, 2026 at $0.50/$2.50 Standard and $1.50/$7.50 Fast, with cache reads at $0.20 and $0.35. Composer 2.5 replaced it as Cursor's most capable in-house model two months later. If you pinned Composer 2 in a config, verify what those requests bill at today rather than assuming the March rates still hold.
Does Cursor give free credits for Composer 2.5?
No dedicated Cursor startup or Composer-specific credit program is published. The credits that cut AI coding costs come from the model and cloud vendors underneath the editor, and a startup can route frontier model calls through its own vendor keys funded by those programs.
The cheapest tier in Cursor is one setting away. The credits are at getaiperks.com.