LatentMD - Content Correct: leaderboard
LatentMD Markdown boundary failures: the share of responses whose content passes all checks on the main prompt grid. Higher is better. Source: arxiv.org/abs/2609.06993.
Metric: Content-correct responses (%). Source: arxiv.org. Saturation forecast: Estimated already saturated. 10 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Claude Sonnet 4 | 93 | #207 |
| 2 | GPT-4o | 80 | #357 |
| 3 | Gemini 2.5 Flash | 76.1 | #255 |
| 4 | Gemma 3 27B | 76 | #606 |
| 5 | Qwen 3 32B | 71.6 | #449 |
| 6 | Llama 3.1 70B | 69.1 | #588 |
| 7 | Llama 3.1 8B | 48.2 | #1130 |
| 8 | Gemma 2 9B | 15.5 | #924 |
No result here: #3 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=latentmd-content-correct · How It Works · Data refreshed daily, snapshot 2026-11-05.