Cursor takes on OpenAI and Anthropic with Composer 2, a code-only model built to match rivals at a fraction of the cost

2026-03-19 · Source: The Decoder · Field: Technology & Digital — Artificial Intelligence & Machine Learning, Software Development & Engineering, Emerging Technologies & Innovation · Depth: Intermediate, short

Summary

Cursor has released Composer 2, its second-generation AI model specifically designed for software development, aiming to compete with leading models from Anthropic and OpenAI at a significantly lower cost. Composer 2 is available in Cursor and the new "Glass" interface, with standard pricing at $0.50 per million input tokens and $2.50 per million output tokens, and a faster variant at $1.50 and $7.50, respectively. This pricing is substantially cheaper than Claude Opus 4.6 ($5.00/$25.00) and GPT-5.4 ($2.50/$15.00). The model achieves a score of 61.3 on Cursor's internal CursorBench, an improvement over Composer 1.5 (44.2) and competitive with Claude Opus 4.6 (58.2) and GPT-5.4 Thinking (63.9). This strategic move allows Cursor to reduce its reliance on third-party models and gain pricing flexibility.

Key takeaway

For CTOs and VPs of Engineering evaluating AI coding assistants, Composer 2 presents a compelling option to significantly reduce operational costs while maintaining competitive performance. Your teams can achieve similar coding assistance capabilities to more expensive models from OpenAI and Anthropic, potentially turning negative margins on consumer subscriptions into profitable ventures. Consider integrating Composer 2 to optimize your AI development budget and enhance your platform's independence.

Key insights

Specialized, cost-effective AI models can compete with generalist leaders by focusing on narrow domains.

Principles

• Domain-specific training enables smaller, cheaper models.
• Reinforcement learning on long-horizon tasks improves code quality.

Method

Composer 2's quality gains stem from stronger continued pretraining followed by reinforcement learning on long-horizon coding tasks, which involve hundreds of actions to complete programming challenges.

In practice

• Consider code-only models for cost-efficient development.
• Evaluate specialized benchmarks like CursorBench for coding AI.

Topics

• Code Generation Models
• AI Model Pricing
• Reinforcement Learning
• AI Benchmarking
• Software Development AI

Best for: CTO, VP of Engineering/Data, Director of AI/ML, Software Engineer, AI Engineer, AI Product Manager

Related on AIssential

• Cursor’s new coding model Composer 2 is here: It beats Claude Opus 4.6 but still trails GPT-5.4 — Cursor's Composer 2 offers a cost-effective, deeply integrated AI coding model optimized for long-horizon agentic workflows.
• Introducing Composer 2 — Composer 2 offers frontier-level coding intelligence with improved performance and competitive pricing through continued pretraining and reinforcement learning.
• Cursor just beat EVERYONE. — Price-performance in coding AI is critical, with workhorse models like Composer 2.5 offering near-frontier capability at a fraction of the cost.
• Cursor Composer 2.5: The AI coding agent that punches way above it’s price tag — High-performance AI coding agents can be achieved at significantly lower costs by leveraging existing powerful open-source models.
• Anthropic just eclipsed OpenAI — Anthropic's new Opus 4.8 model and \$965B valuation mark a significant shift in the AI frontier model landscape.
• A technical report on Composer 2 — Continued pretraining and large-scale RL on realistic data significantly enhance coding model performance and efficiency.

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by The Decoder.