multimodal benchmark
CharXiv-D Leaderboard
CharXiv-D is the descriptive questions subset of the CharXiv benchmark for assessing multimodal large language models on basic information extraction from scientific charts.
Updated Sep 5, 2026
Models17
Model coverage17
MetricScore
EvidenceB
CharXiv-D Ranking
Higher score ranks better on this benchmark.
Sort by
Rank | Model | Score | Percentile | Participants | Evidence | Evaluated |
|---|
| Rank01 | ModelBY | Score95.50% | Percentile100.00% | Participants17 | EvidenceC | Evaluated |
| Rank02 | ModelBY | Score94.60% | Percentile93.75% | Participants17 | EvidenceC | Evaluated |
| Rank03 | ModelAC | Score90.50% | Percentile87.50% | Participants17 | EvidenceC | Evaluated |
| Rank04 | ModelAC | Score90.20% | Percentile81.25% | Participants17 | EvidenceC | Evaluated |
| Rank05 | ModelOP | Score90.00% | Percentile75.00% | Participants17 | EvidenceC | Evaluated |
| Rank06 | ModelOP | Score88.40% | Percentile68.75% | Participants17 | EvidenceC | Evaluated |
| Rank07 | ModelCO | Score88.00% | Percentile62.50% | Participants17 | EvidenceC | Evaluated |
| Rank08 | ModelOP | Score87.90% | Percentile56.25% | Participants17 | EvidenceC | Evaluated |
| Rank09 | ModelAC | Score86.90% | Percentile50.00% | Participants17 | EvidenceC | Evaluated |
| Rank10 | ModelAC | Score85.90% | Percentile43.75% | Participants17 | EvidenceC | Evaluated |
| Rank11 | ModelAC | Score85.50% | Percentile37.50% | Participants17 | EvidenceC | Evaluated |
| Rank12 | ModelOP | Score85.30% | Percentile31.25% | Participants17 | EvidenceC | Evaluated |
| Rank13 | ModelAC | Score83.90% | Percentile25.00% | Participants17 | EvidenceC | Evaluated |
| Rank14 | ModelAC | Score83.00% | Percentile18.75% | Participants17 | EvidenceC | Evaluated |
| Rank15 | ModelAC | Score76.20% | Percentile12.50% | Participants17 | EvidenceC | Evaluated |
| Rank16 | ModelOP | Score73.90% | Percentile6.25% | Participants17 | EvidenceC | Evaluated |
| Rank17 | ModelCO | Score60.00% | Percentile0.00% | Participants17 | EvidenceC | Evaluated |
CharXiv-D Highlights
The leading models and scores on this benchmark.
CharXiv-D Score Distribution
A closer view of the leading scores on this benchmark.
The Top AI Models for CharXiv-D
The first five results on this benchmark, with official price and output speed added where the model identity can be matched.
What is CharXiv-D?
What CharXiv-D measures and how its scores work.
CharXiv-D contains descriptive questions across 2,323 diverse charts from arXiv papers, curated and verified by human experts.
It measures performance on information extraction, enumeration, pattern recognition, and counting questions about scientific charts, reported as Score in ratio units.
Scores are shown in ratio. This benchmark is not independently verified and has an evidence level of B.
- Family
- CharXiv-D
- Modality
- multimodal
- Primary category
- multimodal
- Score direction
- higher
- LLMBoard eligible
- No
- Evaluation key
- overall
LLM Stats. Benchmark scores retain their original unit. Overall score eligibility is shown separately.
FAQ
Common questions about CharXiv-D.
Which model scores highest on CharXiv-D?
Seed 2.1 Pro is currently ranked first with 95.50%.
What are the top three models on CharXiv-D?
The current leaders are Seed 2.1 Pro (95.50%), Seed 2.1 Turbo (94.60%), and Qwen3 VL 32B Instruct (90.50%).
Which CharXiv-D model has the lowest official input price?
GPT-4.1 nano has the lowest matched official input price at $0.10 input / $0.40 output per 1M tokens.
Which models are fastest among CharXiv-D results?
The fastest matched records are GPT-4.1 mini (467.39 tok/s via OpenAI), GPT-4.1 nano (138.14 tok/s via OpenAI), and GPT-4o (132.00 tok/s via OpenAI).
Does the highest score result prove overall model quality?
No. This benchmark measures one defined capability or task. The overall LLMBoard score uses a separate aggregation across eligible benchmark evidence.