Composer 2

Name: Composer 2
Author: Cursor

CursorProprietaryPending Human Review

Composer 2 is Cursor’s Kimi K2.5-derived agentic coding model, released after continued pretraining on code-focused data and large-scale reinforcement learning in realistic Cursor sessions. The training combines stronger base knowledge with the same tools and agent harness used for deployment. It supports multi-step repository work and long-running coding sessions, with a documented 200K single-call context and a faster serving option within the same model identity. The launch on March 19 predates the public technical report. Its ancestry is explicitly documented; this does not make the proprietary post-trained weights publicly downloadable. Performance and cost claims depend on task sets, harnesses, and serving configuration.

2026-03-19

Kimi K2.5-derived Mixture-of-Experts Transformer

Proprietary

Compare with other models

Specifications

Architecture: Kimi K2.5-derived Mixture-of-Experts Transformer
License: Proprietary
Context Window: 200,000 tokens
Type: text
Modalities: text

Benchmark Scores

Advanced Specifications

Model Family: Composer
Finetuned From: Kimi K2.5
API Access: Not Available
Chat Interface: Not Available
Variants: StandardFast serving

Capabilities & Limitations

Capabilities: agentic codingtool uselong-horizon repository workself-summarization
Known Limitations: Public ancestry does not imply public Composer weightsTotal session usage can exceed 200K across repeated calls without a larger single-call windowNo standalone model inference endpoint verifiedCurrent Composer 2 documentation redirects to 2.5; continued Composer 2 serving is not independently verified
Notable Use Cases: software implementationrepository maintenanceinteractive coding agents
Function Calling Support: Yes
Tool Use Support: Yes

Resources

Related Models

Composer 2.5

Cursor

Composer 2.5 is Cursor’s newer Kimi K2.5-derived coding model, significantly improving sustained agent work, tool selection, instruction following, and effort calibration. Training scales synthetic reinforcement-learning tasks and adds localized textual feedback that teaches desired behavior through on-policy distillation. The model retains a documented 200K context window, with standard and fast serving options representing the same intelligence tier. It is available within Cursor’s desktop, web, CLI, and mobile surfaces; the provider also describes building agents with its SDK. The larger model being trained with SpaceXAI is a future project, not a released Composer 3 identity. Composer 2.5 weights remain proprietary.

Composer 1.5

Cursor

Composer 1.5 is Cursor’s significant coding-model update, built by expanding reinforcement learning approximately twentyfold on the same pretrained base used for Composer 1. It adds adaptive thinking and trained self-summarization, balancing fast responses to simple requests with longer reasoning on difficult repository tasks. When a session approaches its context limit, the model writes a summary and continues from that state; this extends task duration rather than the raw context capacity of one inference call. The proprietary mixture-of-experts model is delivered through Cursor. The publisher does not identify the pretrained checkpoint or parameter count, so this record does not invent an ancestry relationship to Composer 1’s post-trained weights.

Composer 1

Cursor

Composer 1 is Cursor’s first publicly released agentic coding language model, introduced alongside Cursor 2.0. The mixture-of-experts model is reinforced on realistic software-engineering problems with the same file-editing, terminal, and semantic-search tools available in the production agent. Low-precision training and coordinated serving target rapid interactive coding. Composer can plan, edit, and answer codebase questions, rather than merely complete lines of code. Its exact parameter count and pretrained ancestry were not disclosed. Access is through Cursor, with no public weights or verified standalone inference endpoint; internal speed and quality comparisons depend on Cursor’s harness and evaluation setup.