image to text benchmark
TextVQA Leaderboard
TextVQA contains 45,336 questions on 28,408 images that require reasoning about text to answer.
Updated Sep 7, 2026
Models16
Model coverage16
MetricScore
EvidenceB
TextVQA Ranking
Higher score ranks better on this benchmark.
Sort by
Rank | Model | Score | Percentile | Participants | Evidence | Evaluated |
|---|
| Rank01 | ModelAC | Score85.50% | Percentile100.00% | Participants16 | EvidenceC | Evaluated |
| Rank02 | ModelAC | Score84.90% | Percentile93.33% | Participants16 | EvidenceC | Evaluated |
| Rank03 | ModelAC | Score84.40% | Percentile86.67% | Participants16 | EvidenceC | Evaluated |
| Rank04 | ModelLA | Score84.30% | Percentile80.00% | Participants16 | EvidenceC | Evaluated |
| Rank05 | ModelDE | Score84.20% | Percentile73.33% | Participants16 | EvidenceC | Evaluated |
| Rank06 | ModelDE | Score83.40% | Percentile66.67% | Participants16 | EvidenceC | Evaluated |
| Rank07 | ModelAM Nova ProAmazon | Score81.50% | Percentile60.00% | Participants16 | EvidenceC | Evaluated |
| Rank08 | ModelDE | Score80.70% | Percentile53.33% | Participants16 | EvidenceC | Evaluated |
| Rank09 | ModelAMNova LiteAmazon | Score80.20% | Percentile46.67% | Participants16 | EvidenceC | Evaluated |
| Rank10 | ModelXA | Score78.10% | Percentile40.00% | Participants16 | EvidenceC | Evaluated |
| Rank11 | ModelMI | Score75.60% | Percentile33.33% | Participants16 | EvidenceC | Evaluated |
| Rank12 | ModelME | Score73.50% | Percentile26.67% | Participants16 | EvidenceC | Evaluated |
| Rank13 | ModelMI | Score72.00% | Percentile20.00% | Participants16 | EvidenceC | Evaluated |
| Rank14 | ModelGO | Score67.70% | Percentile13.33% | Participants16 | EvidenceC | Evaluated |
| Rank15 | ModelGO | Score65.10% | Percentile6.67% | Participants16 | EvidenceC | Evaluated |
| Rank16 | ModelGO | Score57.80% | Percentile0.00% | Participants16 | EvidenceC | Evaluated |
TextVQA Highlights
The leading models and scores on this benchmark.
TextVQA Score Distribution
A closer view of the leading scores on this benchmark.
The Top AI Models for TextVQA
The first five results on this benchmark, with official price and output speed added where the model identity can be matched.
What is TextVQA?
What TextVQA measures and how its scores work.
TextVQA is a benchmark for visual question answering models that focuses on reading and reasoning about text within images.
It measures VQA models' ability to read and reason about text within images, reported using Score as a ratio.
Scores are shown in ratio. This benchmark is not independently verified and has an evidence level of B.
- Family
- TextVQA
- Modality
- multimodal
- Primary category
- image to text
- 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 TextVQA.
Which model scores highest on TextVQA?
Qwen2-VL-72B-Instruct is currently ranked first with 85.50%.
What are the top three models on TextVQA?
The current leaders are Qwen2-VL-72B-Instruct (85.50%), Qwen2.5 VL 7B Instruct (84.90%), and Qwen2.5-Omni-7B (84.40%).
Which TextVQA model has the lowest official input price?
Nova Lite has the lowest matched official input price at $0.06 input / $0.24 output per 1M tokens.
Which models are fastest among TextVQA results?
The fastest matched records are Gemma 3 12B (33.00 tok/s via DeepInfra), Gemma 3 27B (33.00 tok/s via DeepInfra), and Gemma 3 4B (33.00 tok/s via DeepInfra).
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.