Open source · Apache 2.0 · 1.5B params

100% Vectorized Synthetic Data.
Zero Friction.

The first 1.5B parameter code compiler fine-tuned exclusively to write flawless, execution-safe NumPy and Pandas synthetic data scripts. Run locally. Generate 300k rows in < 2 seconds.

View on Hugging FaceGitHub
gnoril — python
>>>

Every model, dataset, and safety guarantee shipped in the open. No accounts, no keys, no rate limits.

Benchmark

Gemini 2.5 Pro vs Gnoril-1.5B

Generating 300,000 rows of validated tabular data.

Gemini 2.5 Protimeout @ 100k · ~$50
Gnoril-1.5B300k rows · 1.2s · free / local

1.2s

Runtime

$0

Cost

300k

Rows

Dataset

The synth-data-10k Dataset

Fine-tuned on 10,000 mathematically verified, AST-validated domain schemas.

{
  "distribution": "gaussian",
  "mean": 4.2,
  "std": 0.87,
  "correlations": [0.31, -0.12],
  "ast_valid": true
}

AST Execution Security

Zero row-level for-loops. Strict Abstract Syntax Tree validation guarantees safe sandbox execution.

Models

The Model Family

Gnoril-1.5BLocal · EdgeGnoril-8BServer