CursorBench
CursorBench is a benchmark framework developed by Cursor for evaluating AI programming models. It systematically assesses the comprehensive capabilities of large language models on real-world programming tasks, covering scenarios such as code generation, code completion, and debugging. Through a standardized task set, it enables cross-model comparisons, providing developers and researchers with objective references for model performance.
Core Facts
CursorBench is a benchmark designed by the Cursor team specifically to evaluate the performance of AI programming models in real IDE environments
CursorBench simulates developers' daily tasks, including scenarios such as understanding cross-file dependency relationships, bug localization and fixing, and multi-module refactoring operations
Timeline (last 90 days)
In the CursorBench benchmark, DeepSeek V4.1 Flash lands on the cost frontier at every effort tier, with per-task cost ranging from $0.60 to $0.98.
Compared with general code-generation benchmarks (such as HumanEval and SWE-bench), CursorBench is characterized by being close to actual editor workflows and incorporating cost as a first-class metric in its reports.
Because the composition of CursorBench questions and the scoring details are not fully public, score comparisons between models should be interpreted with caution.
On the CursorBench benchmark, Muse Wonder scored 46.1% at default effort.
AI evaluation standards are shifting from academically oriented to product-oriented, and the test sets of CursorBench and Vals AI are both closer to real development workflows.
CursorBench's independent evaluation results closely match the data published by Google.
CursorBench is a vendor-built benchmark, with questions not made public and the difficulty distribution unknown.
Gemini 4 Argon scored 52.8% on CursorBench at default reasoning effort, putting it on the same level as Opus 5.5.
There is a gap between benchmark results and real-world experience, as a model's actual performance varies depending on the project's code style, technology stack, and task type.
CursorBench is a vertical-scenario evaluation benchmark, distinct from general-purpose code generation benchmarks such as HumanEval and MBPP.
3 more timeline events
All Facts (10)
CursorBench is a benchmark designed by the Cursor team specifically to evaluate the performance of AI programming models in real IDE environments
80%VerifiedCursorBench simulates developers' daily tasks, including scenarios such as understanding cross-file dependency relationships, bug localization and fixing, and multi-module refactoring operations
75%UnverifiedCompared with general code-generation benchmarks (such as HumanEval and SWE-bench), CursorBench is characterized by being close to actual editor workflows and incorporating cost as a first-class metric in its reports.
50%UnverifiedIn the CursorBench benchmark, DeepSeek V4.1 Flash lands on the cost frontier at every effort tier, with per-task cost ranging from $0.60 to $0.98.
50%UnverifiedOn the CursorBench benchmark, Muse Wonder scored 46.1% at default effort.
50%UnverifiedBecause the composition of CursorBench questions and the scoring details are not fully public, score comparisons between models should be interpreted with caution.
50%UnverifiedAI evaluation standards are shifting from academically oriented to product-oriented, and the test sets of CursorBench and Vals AI are both closer to real development workflows.
50%UnverifiedCursorBench is a vendor-built benchmark, with questions not made public and the difficulty distribution unknown.
50%UnverifiedCursorBench is a vertical-scenario evaluation benchmark, distinct from general-purpose code generation benchmarks such as HumanEval and MBPP.
50%UnverifiedThere is a gap between benchmark results and real-world experience, as a model's actual performance varies depending on the project's code style, technology stack, and task type.
50%