ChartQA
Command A Vision 90.9% leads since Jul 2025 In use Models answer human-written and generated questions about chart images that need visual and arithmetic reasoning. Scored by relaxed accuracy with a 5 percent numeric tolerance.
Top models score near 90 percent, so it is close to saturation.
In the authors’ words
Charts are very popular for analyzing data. When exploring charts, people often ask a variety of complex reasoning questions that involve several logical and arithmetic operations. They also commonly refer to visual features of a chart in their questions. However, most existing datasets do not focus on such complex reasoning questions as their questions are template-based and answers come from a fixed-vocabulary. In this work, we present a large-scale benchmark covering 9.6K human-written questions as well as 23.1K questions generated from human-written chart summaries. To address the unique challenges in our benchmark involving visual and logical reasoning over charts, we present two transformer-based models that combine visual features and the data table of the chart in a unified way to answer questions. While our models achieve the state-of-the-art results on the previous datasets as well as on our benchmark, the evaluation also reveals several challenges in answering complex reasoning questions.
Made by York University and Nanyang Technological University Released Mar 2022 Documents, charts and text in images All tasks public Last in a launch post Aug 2026
Last model added Aug 2026
- Level
- School Students, from grade school to high school
- Scored
- Checked automatically A program checks the result against an answer key, unit tests or the expected final state.
- Answers
- Short answer A number, word or short string matched to a key.
37
Models scored in lab cards
30
Launch posts list it, from 13 labs
23
System cards report it, from 10 labs
Command A Vision leads with 90.9% among models released by Aug 2026
Best score per model as labs report them in their own cards. Labs pick their own settings, so these compare less cleanly than independent runs.
The newest model with a score here came out in Aug 2026, and nobody has reported a newer one. Among the 20 highest-ranked models on the capability index, these have no score yet:
- GPT-6 Astra
- Claude Opus 5.5
- GPT-6.1 Terra
- GPT-6.1 Sol
- and 16 more
- 90.9 Command A Vision
- 90.8 Claude 3.5 Sonnet
- 90.8 Claude 3.5 Sonnet (October 2024)
- 90.3
Qwen3-VL-235B-A22B - 90.3
Nemotron 3 Nano Omni 30B-A3B - 90.0
Llama 4 Maverick - 89.7 Nemotron Nano 12B v2 VL
- 89.5 Qwen2.5-VL-72B
- 89.1
Seed1.5-VL - 89.1 Seed1.6-Thinking
- 88.8 Llama 4 Scout
- 88.0
InternVL3.5-241B-A28B - 87.8
ERNIE 5.0 - 87.6
LongCat-Flash-Omni - 87.4
Mistral Small 3.2 - 87.2
Gemini 1.5 Pro - 87.1 ERNIE-4.5-VL-28B-A3B-Thinking
- 86.8 Qwen3-Omni-30B-A3B
- 86.3 Llama-3.1-Nemotron-Nano-VL-8B-V1
- 86.2 Mistral Small 3.1
- 85.7 GPT-4o
- 85.3 Qwen2.5-Omni-7B
- 83.3
Phi-4-Reasoning-Vision-15B - 82.6 Mistral Medium 3
- 81.8 Phi-3.5-vision-instruct
- 81.7 Claude 3 Haiku
- 81.4 Phi-4-multimodal
- 81.3 Gemini 1.5 Pro
- 81.3 LFM2.5-VL-3B
- 81.1 Claude 3 Sonnet
- 80.8 Gemini 1.0 Ultra
- 80.8 Claude 3 Opus
- 80.8 North Micro Vision Instruct
- 78.5 GPT-4
- 78.0 Gemma 3 27B
- 75.7 Gemma 3 12B
- 68.8 Gemma 3 4B
- 1 Command A Vision 90.9%
- 2 Claude 3.5 Sonnet 90.8%
- 3 Claude 3.5 Sonnet (October 2024) 90.8%
- 4 Qwen3-VL-235B-A22B 90.3%
- 5 Nemotron 3 Nano Omni 30B-A3B 90.3%
- 6 Llama 4 Maverick 90%
- 7 Nemotron Nano 12B v2 VL 89.7%
- 8 Qwen2.5-VL-72B 89.5%
- 9 Seed1.5-VL 89.1%
- 10 Seed1.6-Thinking 89.1%
- 11 Llama 4 Scout 88.8%
- 12 InternVL3.5-241B-A28B 88%
- 13 ERNIE 5.0 87.8%
- 14 LongCat-Flash-Omni 87.6%
- 15 Mistral Small 3.2 87.4%
- 16 Gemini 1.5 Pro May 2024 87.2%
- 17 ERNIE-4.5-VL-28B-A3B-Thinking 87.1%
- 18 Qwen3-Omni-30B-A3B 86.8%
- 19 Llama-3.1-Nemotron-Nano-VL-8B-V1 86.3%
- 20 Mistral Small 3.1 86.2%
- 21 GPT-4o May 2024 85.7%
- 22 Qwen2.5-Omni-7B 85.3%
- 23 Phi-4-Reasoning-Vision-15B 83.3%
- 24 Mistral Medium 3 82.6%
- 25 Phi-3.5-vision-instruct 81.8%
- 26 Claude 3 Haiku 81.7%
- 27 Phi-4-multimodal 81.4%
- 28 Gemini 1.5 Pro Feb 2024 81.3%
- 29 LFM2.5-VL-3B 81.3%
- 30 Claude 3 Sonnet 81.1%
- 31 Gemini 1.0 Ultra 80.8%
- 32 Claude 3 Opus 80.8%
- 33 North Micro Vision Instruct 80.8%
- 34 GPT-4 Mar 2023 78.5%
- 35 Gemma 3 27B 78%
- 36 Gemma 3 12B 75.7%
- 37 Gemma 3 4B 68.8%
| # | Model | Score | Setting | Date | Source |
|---|---|---|---|---|---|
| 1 | Command A Vision | 90.9% | Jul 2025 | Model card | |
| 2 | Claude 3.5 Sonnet | 90.8% | Jun 2024 | Model card | |
| 3 | Claude 3.5 Sonnet (October 2024) | 90.8% | Oct 2024 | Model card | |
| 4 | Qwen3-VL-235B-A22B | 90.3% | Nov 2025 | Model card | |
| 5 | Nemotron 3 Nano Omni 30B-A3B | 90.3% | Apr 2026 | Model card | |
| 6 | Llama 4 Maverick | 90% | Apr 2025 | Model card | |
| 7 | Nemotron Nano 12B v2 VL | 89.7% | Oct 2025 | Model card | |
| 8 | Qwen2.5-VL-72B | 89.5% | Jan 2025 | Model card | |
| 9 | Seed1.5-VL | 89.1% | May 2025 | Model card | |
| 10 | Seed1.6-Thinking | 89.1% | Jun 2025 | Model card | |
| 11 | Llama 4 Scout | 88.8% | Apr 2025 | Model card | |
| 12 | InternVL3.5-241B-A28B | 88% | Aug 2025 | Model card | |
| 13 | ERNIE 5.0 | 87.8% | Feb 2026 | Model card | |
| 14 | LongCat-Flash-Omni | 87.6% | Oct 2025 | Model card | |
| 15 | Mistral Small 3.2 | 87.4% | Jun 2025 | Model card | |
| 16 | Gemini 1.5 Pro May 2024 | 87.2% | May 2024 | Model card | |
| 17 | ERNIE-4.5-VL-28B-A3B-Thinking | 87.1% | Nov 2025 | Model card | |
| 18 | Qwen3-Omni-30B-A3B | 86.8% | Sep 2025 | Model card | |
| 19 | Llama-3.1-Nemotron-Nano-VL-8B-V1 | 86.3% | Jun 2025 | Model card | |
| 20 | Mistral Small 3.1 | 86.2% | Mar 2025 | Model card | |
| 21 | GPT-4o May 2024 | 85.7% | May 2024 | Model card | |
| 22 | Qwen2.5-Omni-7B | 85.3% | Mar 2025 | Model card | |
| 23 | Phi-4-Reasoning-Vision-15B | 83.3% | Mar 2026 | Model card | |
| 24 | Mistral Medium 3 | 82.6% | May 2025 | Model card | |
| 25 | Phi-3.5-vision-instruct | 81.8% | Aug 2024 | Model card | |
| 26 | Claude 3 Haiku | 81.7% | Mar 2024 | Model card | |
| 27 | Phi-4-multimodal | 81.4% | Feb 2025 | Model card | |
| 28 | Gemini 1.5 Pro Feb 2024 | 81.3% | Feb 2024 | Model card | |
| 29 | LFM2.5-VL-3B | 81.3% | Aug 2026 | Model card | |
| 30 | Claude 3 Sonnet | 81.1% | Mar 2024 | Model card | |
| 31 | Gemini 1.0 Ultra | 80.8% | Dec 2023 | Model card | |
| 32 | Claude 3 Opus | 80.8% | Mar 2024 | Model card | |
| 33 | North Micro Vision Instruct | 80.8% | Aug 2026 | Model card | |
| 34 | GPT-4 Mar 2023 | 78.5% | Mar 2023 | Model card | |
| 35 | Gemma 3 27B | 78% | Mar 2025 | Model card | |
| 36 | Gemma 3 12B | 75.7% | Mar 2025 | Model card | |
| 37 | Gemma 3 4B | 68.8% | Mar 2025 | Model card |
8 models have held number one since 2023
Lab claims count from the day the card came out, independent runs from the model’s release.
OpenAI Google DeepMind Anthropic Cohere
The best score passed 85% in May 2024
All the scores we have, one dot each, so a model tested at several settings appears more than once.
Reported by the lab in its model card
Best so far
2 labs put it in a launch post in the last six months
Launch posts from every lab that has listed it, one tick per post. Tall colored ticks include it, and short gray ones leave it out.
| Date | Launch post | Score | As the post writes it |
|---|---|---|---|
| 12 Aug 2026 | LFM2.5-VL-3B | 81.3 | ChartQA (test) |
| 12 Aug 2026 | North Micro Vision Instruct | 80.8 | ChartQA Test |
| 4 Mar 2026 | Phi-4-Reasoning-Vision-15B | 83.3 | ChartQA_TEST (default mode) |
| 6 Feb 2026 | ERNIE 5.0 | 87.8 | ChartQA |
| 11 Nov 2025 | ERNIE-4.5-VL-28B-A3B-Thinking | 87.1 | ChartQA |
| 31 Oct 2025 | LongCat-Flash-Omni | 87.6 | ChartQA |
| 28 Oct 2025 | Nemotron Nano 12B v2 VL | 89.7 | ChartQA (test) |
| 22 Sep 2025 | Qwen3-Omni-30B-A3B | 86.8 | ChartQA_test |
| 31 Jul 2025 | Command A Vision | 90.9 | ChartQA |
| 25 Jun 2025 | Seed1.6-Thinking | 89.1 | ChartQA |
| 20 Jun 2025 | Mistral Small 3.2 | 87.4 | ChartQA |
| 3 Jun 2025 | Llama-3.1-Nemotron-Nano-VL-8B-V1 | 86.3 | ChartQA |
| 12 May 2025 | Seed1.5-VL | 89.1 | ChartQA |
| 7 May 2025 | Mistral Medium 3 | 82.6 | ChartQA 0-shot |
| 5 Apr 2025 | Llama 4 Scout | 88.8 | ChartQA, relaxed_accuracy |
| 5 Apr 2025 | Llama 4 Maverick | 90 | ChartQA, relaxed_accuracy |
| 27 Mar 2025 | Qwen2.5-Omni-7B | 85.3 | ChartQAtest Avg |
| 17 Mar 2025 | Mistral Small 3.1 | 86.2 | ChartQA |
| 16 Mar 2025 | ERNIE 4.5 | ChartQA | |
| 12 Mar 2025 | Gemma 3 27B | 78 | ChartQA (IT); PT = 76.3 |
| 26 Feb 2025 | Phi-4-multimodal | 81.4 | ChartQA |
| 28 Jan 2025 | Qwen2.5-VL-72B | 89.5 | ChartQA_TEST |
| 22 Oct 2024 | Claude 3.5 Sonnet (October 2024) | 90.8 | ChartQA (test, relaxed accuracy) |
| 20 Jun 2024 | Claude 3.5 Sonnet | 90.8 | ChartQA (test, relaxed accuracy) |
| 14 May 2024 | Gemini 1.5 Pro | 87.2 | ChartQA (test), 0-shot |
| 13 May 2024 | GPT-4o | 85.7 | ChartQA (%) (test) |
| 4 Mar 2024 | Claude 3 Opus | 80.8 | ChartQA (test, relaxed accuracy) CoT |
| 15 Feb 2024 | Gemini 1.5 Pro | 81.3 | ChartQA (test), 0-shot |
| 6 Dec 2023 | Gemini 1.0 Ultra | 80.8 | ChartQA (test) |
| 14 Mar 2023 | GPT-4 | 78.5 | ChartQA Relaxed accuracy (test) |
23 system cards report it, from 10 labs
Besides launch posts, we read the system cards, model cards and technical reports labs publish with a release. They report far more, including safety and frontier-risk tests.
| Date | Document | Score | As the document writes it | Section |
|---|---|---|---|---|
| 27 Apr 2026 | Nemotron 3 Nano Omni 30B-A3B technical report | 90.3 | ChartQA (Test), reasoning on, 89.9 reasoning off (Table 7) | Capabilities |
| 4 Mar 2026 | Phi-4-Reasoning-Vision-15B technical report | 83.3 | ChartQA TEST, default mode, 76.5 force nothink, 82.9 force thinking (Tables 5 and 6) | Capabilities |
| 4 Feb 2026 | ERNIE 5.0 technical report | 87.8 | ChartQA (Table 3) | Capabilities |
| 26 Nov 2025 | Qwen3-VL-235B-A22B technical report | 90.3 | ChartQA test (Table 2) | Capabilities |
| 6 Nov 2025 | Nemotron Nano 12B v2 VL technical report | 84.9 | ChartQA (Test), reasoning on (Table 3). Reasoning off 89.8 | Capabilities |
| 31 Oct 2025 | LongCat-Flash-Omni technical report | 87.6 | ChartQA (Table 5) | Capabilities |
| 22 Sep 2025 | Qwen3-Omni-30B-A3B technical report | 86.8 | ChartQA test Avg., Instruct (Table 9). Thinking 89.5 (Table 10) | Capabilities |
| 25 Aug 2025 | InternVL3.5-241B-A28B technical report | 88 | ChartQA (test avg) (Table 4) | Capabilities |
| 11 May 2025 | Seed1.5-VL technical report | 89.1 | ChartQA, thinking (Table 6). Non-thinking 87.4 | Capabilities |
| 5 Apr 2025 | Llama 4 Scout model card | 88.8 | ChartQA, 0-shot, relaxed_accuracy, instruction tuned (Instruction tuned models table) | Capabilities |
| 5 Apr 2025 | Llama 4 Maverick model card | 90 | ChartQA, 0-shot, relaxed_accuracy, instruction tuned (Instruction tuned models table) | Capabilities |
| 26 Mar 2025 | Qwen2.5-Omni-7B technical report | 85.3 | ChartQA test Avg (Table 5) | Capabilities |
| 25 Mar 2025 | Gemma 3 4B technical report | 68.8 | ChartQA, IT, P&S applied (Table 16) | Capabilities |
| 25 Mar 2025 | Gemma 3 27B technical report | 78 | ChartQA, IT, P&S applied (Table 16) | Capabilities |
| 25 Mar 2025 | Gemma 3 12B technical report | 75.7 | ChartQA, IT, P&S applied (Table 16) | Capabilities |
| 3 Mar 2025 | Phi-4-multimodal technical report | 81.4 | ChartQA (test) (Table 1). Speech-prompt ShareGPT4o ChartQA 69.0 (Table 2) | Capabilities |
| 19 Feb 2025 | Qwen2.5-VL-72B technical report | 89.5 | ChartQA test Avg. (Table 5) | Capabilities |
| 22 Oct 2024 | Claude 3.5 Sonnet (October 2024) model card | 90.8 | ChartQA (test, relaxed accuracy), 0-shot CoT (Table 5) | Capabilities |
| 30 Aug 2024 | Phi-3.5-vision-instruct technical report | 81.8 | ChartQA (test) (Table 5) | Capabilities |
| 20 Jun 2024 | Claude 3.5 Sonnet model card | 90.8 | ChartQA (test, relaxed accuracy), 0-shot CoT (Table 2) | Capabilities |
| 4 Mar 2024 | Claude 3 Sonnet model card | 81.1 | ChartQA (test, relaxed accuracy), 0-shot CoT (Table 3) | Capabilities |
| 4 Mar 2024 | Claude 3 Opus model card | 80.8 | ChartQA (test, relaxed accuracy), 0-shot CoT (Table 3) | Capabilities |
| 4 Mar 2024 | Claude 3 Haiku model card | 81.7 | ChartQA (test, relaxed accuracy), 0-shot CoT (Table 3) | Capabilities |
Versions
More in Vision
- MMMU In use 11.5K college-level multimodal questions across 30 subjects. Multiple-choice and open accuracy. GPT-5.1 85.4%
- MMMU-Pro In use Harder MMMU: college-level multimodal questions with more options and vision-only inputs to remove text shortcuts. Kimi K3.1 Pro 84.1%
- CharXiv In use Chart understanding on real arXiv figures. Reasoning questions require synthesizing information from complex charts. Claude Fable 5 93.5%
- MATH-Vision (MathVision) In use 3,040 competition math problems with visual context. Answer accuracy. Qwen 3.8 Max 97.7%
- DocVQA In use Models answer 50,000 questions about 12,000 scanned document images by reading their text and layout. Scoring is ANLS, a string similarity to reference answers. Seed1.5-VL 96.9%
- MMStar In use Models answer 1,500 human-curated image questions that cannot be solved without the image, across 6 core capabilities. Scored by accuracy. Qwen 3.8 Max 85.9%