Composer 2.5: Pricing, Fast Mode and Uses

Composer 2.5The current Composer release, better at long-running tasks and at judging when a job needs a light touch versus deep work. Press Enter for the full definition. is Cursor's own coding model, trained for long tasks that use files, terminal commands and other agent tools. Cursor rates its speed fast, its cost low and its intelligence frontier. Over Composer 2Cursor's own coding model, built on the open Kimi K2.5 base as a software-engineering specialist. Cursor calls it frontier-level at coding and more cost-efficient than comparable models; Composer 2.5 has since replaced it in Cursor's model list. Press Enter for the full definition., the release improved sustained work, complex instruction following, effort calibration, tool selection, intent understanding, communication and reliability.

Its context window is 200k, and the model card lists no larger maximum. That is smaller than the 1M ceilings on some third-party frontier models, but context size is only one part of task performance. Cursor tuned 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 tool use, file edits and terminal operations inside Cursor, which matters most when a task needs many tool calls and the per-token cost adds up.

Use Fast for an interactive pairing loop where response time keeps a person moving. Use Standard for queued cloud work, high-volume batches, or any run where nobody waits on each tool result. Cursor says the two variants have the same intelligence. They differ in speed and price.

Fast is the product default, but only on plans that offer it: Plus and legacy Cursor StartAn India-only legacy individual plan at ₹649 a month centered on Cursor's own models, with cloud and mobile access but without Other Models, Auto, Bugbot, Automations or the Agent SDK; Cursor names Plus as the comparable plan on The X Subscription. Press Enter for the full definition. run Composer 2.5The current Composer release, better at long-running tasks and at judging when a job needs a light touch versus deep work. Press Enter for the full definition. at Standard.

Fast costs six times the Standard input and output rates, with cache reads at $0.50 instead of $0.20. That premium can pay off in a short feedback loop, and it looks wasteful on a background migration that takes time either way. Check the selected variant explicitly, because the plain model name can resolve to the product default on some surfaces.

Run one comparison on the same task before you change a team default. Record elapsed time, usage, final checks and human waiting time. Fast can be worth more than its token premium if it keeps an expensive review session moving. Standard can come out ahead when an agent runs unattended overnight and nobody is waiting.

Cursor describes a substantial improvement in intelligence and behavior. The model stays with long-running work more reliably, follows complex instructions better, calibrates effort to task difficulty and selects tools more carefully, and the changelog calls it more pleasant to collaborate with. Behavior matters here because a long task can fail on one confused tool call or one lost constraint, even when its benchmark score looks strong.

Cursor's research post explains part of the training change. Composer 2.5The current Composer release, better at long-running tasks and at judging when a job needs a light touch versus deep work. Press Enter for the full definition. kept the same open checkpoint as Composer 2Cursor's own coding model, built on the open Kimi K2.5 base as a software-engineering specialist. Cursor calls it frontier-level at coding and more cost-efficient than comparable models; Composer 2.5 has since replaced it in Cursor's model list. Press Enter for the full definition., Moonshot's Kimi K2.5Moonshot's open model that Cursor used as the base for Composer 2 and Composer 2.5 (1T parameters, 32B active, 256K context). Cursor's report says it chose Kimi K2.5 for strong results on its internal tests and for its efficiency on Cursor's infrastructure. Press Enter for the full definition.. Cursor then scaled its training, built harder reinforcement-learning environments, added targeted textual feedback on specific mistakes, and trained it on 25 times more synthetic tasks than Composer 2. Cursor says existing benchmarks don't capture the communication and effort-calibration work well.

Start with 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 everyday implementation, repetitive repository work and long agent runs that lean on tool use. The model can search the project, edit files, run the terminal, use the browser, search the web, generate images, ask questions and fetch rules, within the permissions of the surface you run it on.

Judge the price by cost per accepted task. A cheap model that needs three retries can cost more than one stronger attempt, and an expensive model is wasted on a task the cheaper one finishes. Keep a small evaluation set from real work, and let it show where 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.'s price advantage survives your quality bar.

Usage comes from the included pool for Cursor's own models, which Composer 2.5The current Composer release, better at long-running tasks and at judging when a job needs a light touch versus deep work. Press Enter for the full definition. shares with Grok 4.7Cursor and SpaceXAI's current lead Cursor model, in the weekly pool at $2 per million input tokens and $6 per million output, with a 256k context window and a 500k variant; above 256k input tokens it bills at 2x the standard rate. Press Enter for the full definition., Grok 4.6A Cursor and SpaceXAI model for complex coding and knowledge work, still in the Cursor model pool and succeeded by Grok 4.7; it is a selectable model and is separate from the Grok Bot product. Press Enter for the full definition. and Grok 4.5. 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. that is the weekly pool every paid plan includes, and on legacy plans it is the monthly Cursor ModelsOn legacy Cursor plans, the monthly usage pool for Cursor's own models (Grok 4.7, Grok 4.6, Grok 4.5 and Composer 2.5), kept separate from the third-party Other Models pool. The X Subscription replaces it with one weekly pool. Press Enter for the full definition. pool. Third-party models sit outside it. The Standard and Fast rates apply to usage beyond the pool.

Your plan decides which variants you can pick: Premium and above, and legacy Pro and higher, let you choose Fast, while Plus and legacy Cursor StartAn India-only legacy individual plan at ₹649 a month centered on Cursor's own models, with cloud and mobile access but without Other Models, Auto, Bugbot, Automations or the Agent SDK; Cursor names Plus as the comparable plan on The X Subscription. Press Enter for the full definition. can't. The Cursor Token Rate, which Teams and Enterprise pay on eligible third-party requests and some Auto routes, never applies to Composer 2.5The current Composer release, better at long-running tasks and at judging when a job needs a light touch versus deep work. Press Enter for the full definition..

You can pick Composer 2.5The current Composer release, better at long-running tasks and at judging when a job needs a light touch versus deep work. Press Enter for the full definition. in the editor and the CLI, and in the Cursor SDK on plans that include it. For Cloud AgentsAgents that run in a Cursor-managed virtual machine, check out the repo, do the work and open a pull request, then shut down, with no load on your laptop. Press Enter for the full definition. on the web and in the mobile app, Cursor's docs say you can choose any model available for cloud agents but don't list those models.

Read your remaining usage in the dashboard instead of translating the list rate into an assumed request allowance. Each task consumes a different amount of input, cached context, output and tool-loop history, and the dashboard shows what your real workload spends.

When you evaluate a new frontier model, 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. makes a useful baseline. Run the cheaper Cursor model first on a task you can verify, and escalate only if the result misses. That keeps the comparison tied to work the team does, and it stops a launch benchmark from deciding every model selection in advance.

For shared workflows, record the variant as well as the model. A report that says Composer 2.5The current Composer release, better at long-running tasks and at judging when a job needs a light touch versus deep work. Press Enter for the full definition. without Standard or Fast leaves out the largest price difference in the choice. Record the surface too, since product defaults and explicit SDK parameters can resolve the same short model ID differently. Rerun the baseline after meaningful model or harness changes, and as the codebase grows. A small, stable evaluation set keeps the evidence current without turning model selection into a permanent testing project.