Programming Language Classification Dataset with 1.66 Million Code Samples

1.664 million cleaned and labeled source code samples across 16 programming languages, curated for language identification tasks. The dataset was created by author kaushik-harsh-99 and is hosted on Hugging Face. It was last updated on May 30, 2026.

Use Cases

• Train a classifier to identify programming languages based on source code snippets.
• Benchmark language identification models against a large-scale, balanced dataset.
• Fine-tune a code model for downstream tasks requiring language context.
• Analyze syntactic patterns across different programming languages.

Strengths

• Contains 1.664 million data points, indicating a substantial scale.
• Explicitly balanced across 16 programming languages, which can help mitigate class imbalance.
• Described as cleaned and specifically curated for classification tasks, suggesting intentional quality control.

Limitations

• Column-level documentation is absent; field semantics must be inferred after download.
• Row count is known, but file size, formats, and specific column names are unknown.
• Last updated 2026-05-30 11:14:52; freshness should be verified.

Provenance

Source

huggingface

Collection Method

Curated from source code snippets; specific sources are not detailed.

Freshness

Last updated May 30, 2026.