ChartQA

ChartQA is a large-scale benchmark comprising 9.6K human-written questions and 23.1K questions generated from human-written chart summaries, designed to evaluate models' abilities in visual and logical reasoning over charts.

ChartQA Leaderboard

AllOpenProprietary

24 models

#	Model	Score	Size	Context	Cost	License
1	Claude 3.5 Sonnet Anthropic	0.908	—	200K	$3.00 / $15.00
2	Llama 4 Maverick Meta	0.900	400B	1.0M	$0.17 / $0.60
3	Qwen2.5 VL 72B Instruct Alibaba Cloud / Qwen Team	0.895	72B	—	—
4	Nova Pro Amazon	0.892	—	300K	$0.80 / $3.20
5	Llama 4 Scout Meta	0.888	109B	10.0M	$0.08 / $0.30
6	Qwen2-VL-72B-Instruct Alibaba Cloud / Qwen Team	0.883	73B	—	—
7	Pixtral Large Mistral AI	0.881	124B	128K	$2.00 / $6.00
8	Mistral Small 3.2 24B Instruct Mistral AI	0.874	24B	—	—
9	Qwen2.5 VL 7B Instruct Alibaba Cloud / Qwen Team	0.873	8B	—	—
10	Nova Lite Amazon	0.868	—	300K	$0.06 / $0.24
11	DeepSeek VL2 DeepSeek	0.860	27B	—	—
12	GPT-4o OpenAI	0.857	—	128K	$2.50 / $10.00
13	Llama 3.2 90B Instruct Meta	0.855	90B	128K	$0.35 / $0.40
14	Qwen2.5-Omni-7B Alibaba Cloud / Qwen Team	0.853	7B	—	—
15	DeepSeek VL2 Small DeepSeek	0.845	16B	—	—
16	Llama 3.2 11B Instruct Meta	0.834	11B	128K	$0.05 / $0.05
17	Pixtral-12B Mistral AI	0.818	12B	128K	$0.15 / $0.15
17	Phi-3.5-vision-instruct Microsoft	0.818	4B	—	—
19	Phi-4-multimodal-instruct Microsoft	0.814	6B	128K	$0.05 / $0.10
20	DeepSeek VL2 Tiny DeepSeek	0.810	3B	—	—
21	Gemma 3 27B Google	0.780	27B	131K	$0.10 / $0.20
22	Grok-1.5V xAI	0.761	—	—	—
23	Gemma 3 12B Google	0.757	12B	131K	$0.05 / $0.10
24	Gemma 3 4B Google	0.688	4B	131K	$0.02 / $0.04