# GDP.pdf task-macro rubric pass rate with Python, Gemini judge (GDP.pdf w/ Python (Gemini judge))

> Professional document understanding where the model receives each original PDF in a sandbox with an IPython tool.

Canonical page: https://benchlm.ai/benchmarks/gdppdfpythongeminijudge

- Category: [Multimodal & Grounded](/multimodal-grounded)
- Last updated: October 29, 2026

## About GDP.pdf w/ Python (Gemini judge)

- Year: 2026
- Tasks: Professional document prompts with Python
- Format: Mean rubric pass rate
- Difficulty: Tool-assisted document reasoning
- Paper: [GDP.pdf](https://surgehq.ai/blog/gdp-pdf-can-100b-ai-models-master-the-documents-that-run-the-world)

Moonshot grades each rubric criterion with Gemini-3.5-Flash at medium effort using the official judging prompt and reports the task-macro mean rubric pass rate over five runs. The judge and tool setup differ from Anthropic's GDP.pdf tools row, so BenchLM keeps this protocol on its own key.

GDP.pdf w/ Python (Gemini judge) is currently displayed on BenchLM for reference, but it is excluded from the weighted scoring formula.

## Leaderboard (1 models)

| Rank | Model | Creator | Score |
|------|-------|---------|-------|
| 1 | [Kimi K3.1 Pro](/models/kimi-3-1-pro) | Moonshot AI | 85.7% |

## FAQ

### What does GDP.pdf w/ Python (Gemini judge) measure?

Professional document understanding where the model receives each original PDF in a sandbox with an IPython tool.

### Which model scores highest on GDP.pdf w/ Python (Gemini judge)?

Kimi K3.1 Pro by Moonshot AI currently leads with a score of 85.7% on GDP.pdf w/ Python (Gemini judge).

### How many models are evaluated on GDP.pdf w/ Python (Gemini judge)?

1 AI models have published results on GDP.pdf w/ Python (Gemini judge) in the BenchLM catalog.

### Does GDP.pdf w/ Python (Gemini judge) affect BenchLM's overall score?

Not directly. GDP.pdf w/ Python (Gemini judge) is still displayed on BenchLM for reference, but it is excluded from the weighted scoring formula.