Moondream3 Preview Release - POPE: leaderboard
Moondream September 18, 2025 preview release, official overview table; exact printed model labels retained. Unstarred source population only. Release marks GPT5/Claude4Sonnet ChartQA and DocVQA cells as 100-random-question samples; those four cells are withheld, never pooled with full unstarred evaluations. Reasoning effort, API snapshots and precision are not specified and are not inferred from model-card defaults. Exact POPE split/aggregation and score statistic are not stated in this table; do not relabel as F1 or accuracy from a separate evaluator. Metric: Source benchmark score. Higher is better. Source: https://docs.moondream.ai/
Metric: Source benchmark score. Source: docs.moondream.ai. Saturation forecast: Estimated already saturated. 4 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | GPT-5 | 88.4 | #95 |
| 2 | Gemini 2.5 Flash | 88.1 | #243 |
| 3 | Claude Sonnet 4 | 84.6 | #196 |
No result here: #3 Claude Opus 5.5, #5 GPT-6 Astra, #7 Gemini 4 Argon.
Interactive version: theaggregate.ai/benchmark?slug=moondream3-preview-release-pope · How It Works · Data refreshed daily, snapshot 2026-10-17.