Cursor Now Supports GPT-5.4, Its Current Benchmark Leader

OpenAI

Mar 5, 2026 · Updated Apr 25, 2026

Cursor added GPT-5.4 to its AI coding editor, ranking it first on Cursor's internal benchmarks. The model is confident and decisive in tackling messy, ambiguous coding problems, and is strong at parallelizing work across long agent sessions.

Cursor, an AI coding editor, added GPT-5.4 — OpenAI's frontier model — to its lineup. It currently leads all other models on Cursor's internal benchmarks. The model carries a 272k context window with 1M token Max Mode support and Agent + Thinking capabilities. Cursor describes it as confident and decisive: it works through ambiguous problems without second-guessing and maintains momentum across long workflows.

For developers running heavy agentic coding sessions, this means the most intelligent model in Cursor's lineup also handles parallelism and sustained momentum well — traits that matter more as sessions grow longer and more complex. For cost-sensitive work, Cursor positions GPT-5.3 Codex as a lower-cost alternative where raw intelligence isn't the priority.

To try it, open Cursor's model selector and switch to GPT-5.4. It draws from the API pool at $2.5/M input and $15/M output. A faster variant, gpt-5.4-fast, runs 15% faster at 2x the standard price.

View the full update on cursor.com

Cursor

@cursor_aiMar 5

GPT 5.4 is now available in Cursor! We've found it to be more natural and assertive than previous models. It's currently the leader on our internal benchmarks.

143

View on X

Every HeadsUpAI update is written based on its original source and reviewed before it's published. Read our editorial standards →

See all AI news & updates from OpenAI →

Keep reading

Cursor Adds GPT-5.5 and Leads Internal Benchmarks With 50 Percent Discount

Cursor integrated OpenAI's GPT-5.5, which now leads its internal CursorBench evaluation with a 72.8% score. The update highlights a shift toward using real-world developer sessions rather than public benchmarks to measure frontier model performance on complex, multi-file coding tasks.

Cursor Publishes CursorBench, Its Internal Agentic Coding Evaluation Methodology

OpenAIMar 15

Cursor Publishes CursorBench, Its Internal Agentic Coding Evaluation Methodology

Cursor published CursorBench, its internal eval suite that scores models on real coding agent tasks from actual developer sessions. Public benchmarks struggle to differentiate frontier models reliably — CursorBench produces more separation where it matters most.

Windsurf Adds GPT-5.4 Mini for Faster Coding and Subagent Workflows

WindsurfMar 20

Windsurf Adds GPT-5.4 Mini for Faster Coding and Subagent Workflows

Windsurf now supports GPT-5.4 mini, OpenAI's model optimized for coding, computer use, multimodal understanding, and subagents. It runs 2x faster than GPT-5 mini, making it a quicker option for developers already working inside the IDE.