ParseBench - Content Faithfulness: leaderboard
LlamaIndex's ParseBench scores document parsing of human-verified enterprise pages from insurance, finance and government documents with rule-based checks; VLM rows only, parser products excluded. This board reports the Content Faithfulness dimension: penalises omitted text, hallucinated text and reading-order violations.
Metric: Content Faithfulness Score. Source: parsebench.ai. Saturation forecast: Around March 2028. 121 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Claude Opus 5.5 | 91.35 | #3 |
| 2 | Claude Opus 5.5 (High) | 91.25 | #3 (Claude Opus 5.5) |
| 3 | Claude Sonnet 5.5 | 91.24 | #13 |
| 4 | Claude Opus 5.5 (Low) | 91.23 | #3 (Claude Opus 5.5) |
| 5 | Claude Fable 5.1 | 90.77 | #10 |
| 6 | Qwen 3.5 9B | 90.5 | #395 |
| 7 | Claude Opus 4.7 | 90.26 | #53 |
| 8 | Claude Fable 5 | 90.02 | #19 |
| 9 | Gemini 3.5 Flash (Medium) | 90 | #68 (Gemini 3.5 Flash) |
| 10 | Gemma 4 31B (IT) | 89.9 | #180 |
| 11 | Claude Haiku 5.5 | 89.83 | #46 |
| 12 | Gemini 3 Flash (High) | 89.74 | #105 (Gemini 3 Flash) |
| 13 | GPT-6.1 Terra (Low) | 89.57 | #14 (GPT-6.1 Terra) |
| 14 | GPT-6.1 Terra (Medium) | 89.32 | #14 (GPT-6.1 Terra) |
| 15 | Gemini 3.8 Flash (High) | 89.31 | #30 (Gemini 3.8 Flash) |
No result here: #9 Gemini 4 Argon.
Interactive version: theaggregate.ai/benchmark?slug=parsebench-content-faithfulness · How It Works · Data refreshed daily, snapshot 2026-11-07.