Composer 2.5

Name: Composer 2.5
Author: Cursor (Anysphere)

Released

2026-05-18

Last refreshed

2026-06-30

Status

Researched 94d ago

ProprietaryCommercial use: conditionalCodingRAGAgentsLong contextJSON / Tool use

Cursor (Anysphere) releases · 2 in the last 12 monthsChangelog →

Specifications

Family: Composer
Released: 2026-05-18
Context: 1m
Specialization: code
Openness: Proprietary
License: ProprietaryCommercial use: conditional
Weights: Not released
Code: Not released

Created by

Cursor (Anysphere)

AI-driven multi-file code generation

Buffalo, New York, United States

Founded 2022

Website

Pricing

Output / 1M

$2.50

Input / 1M

$0.500

Cheapest of 1 route · Cursor

Providers(1)

Cursor

View 1 provider route

Links

Website

About

Cursor's agentic coding model released May 18, 2026. Built on Moonshot AI's Kimi K2.5 open-source checkpoint with targeted RL using textual feedback and 25× more synthetic training tasks than Composer 2. Designed for long-horizon software engineering tasks: multi-file edits, terminal command execution, codebase-wide semantic search, and autonomous task planning. Uses Cursor's compaction-in-the-loop context management for long coding sessions. Available on all Cursor plans; accessed through the Cursor IDE (not a standalone API). Standard pricing: $0.50/M input, $2.50/M output; Fast (default): $3.00/M input, $15.00/M output.

Provider price ladder

Compare API pricing across 1 providers for input and output tokens, batch, and cached reads when available.

Provider	Input / 1M	Output / 1M	Route
Cursor	$0.500	$2.50	Serverless

Capabilities

JSON / Tool useCode Execution

Benchmark peer barsfor Coding

No task-mapped benchmark peers are available for this model yet.

Benchmark scores(6)

Scores are benchmark-specific and are direction-aware: the same numeric gap can mean very different outcomes across suites. Use the leaderboard context and this model's provider route to decide whether the winning margin is meaningful for your workload.

Benchmark	Score	Version	Evaluation	Source
Terminal-Bench 2.0	69.3	Terminal-Bench 2.0Observed 2026-05-21	—	Source
SWE-bench Multilingual	79.8	SWE-bench MultilingualObserved 2026-05-21	—	Source
CursorBench	63.2	CursorBench 3.1Observed 2026-06-30	Configuration: Composer 2.5 (single reported configuration) Harness: CursorBench 3.1 Evaluator: Cursor Confidence: confirmed Notes: Cursor published one CursorBench 3.1 configuration for this model; no cross-effort selection was needed.	Source
SWE-rebench	79.8	Reported as SWE-bench Multilingual (300 tasks, 9 languages) (resolved%)Observed 2026-06-07	—	Source
CursorBench	56.1	CursorBench 3.2Observed 2026-07-18	Configuration: Composer 2.5 Harness: CursorBench 3.2 productized Cursor-agent workflow Evaluator: Cursor Confidence: confirmed Cost/task: $0.44 Tokens/task: 14,286 Steps/task: 33 Notes: Cursor vendor-reported result; not independently reproducible. Results are subject to variance, and small score differences may not be statistically meaningful.	Source
CursorBench	27.7	CursorBench 4.0Observed 2026-09-10	Configuration: Composer 2.5 Harness: CursorBench 4.0 productized Cursor-agent workflow Evaluator: Cursor Confidence: confirmed Cost/task: $0.68 Tokens/task: 17,347 Steps/task: 41 Notes: Cursor vendor-reported configuration result; not independently reproducible. Results are subject to variance, and small score differences may not be statistically meaningful.	Source