AI2D

Overview

Type: percentage

Recorded Results

Average Score

87.35%

Score Range

71.60% - 94.40%

Leading Model (lowest score)

71.60% - DeepSeek VL2 Tiny

Scores Over Time

Individual benchmark scores plotted by date.

Models Using This Benchmark

Organisation	Model	Reported	Top Score	Info	Self Reported	Source
DeepSeek	DeepSeek VL2 Tiny	13 Dec 2024	71.60%	-	Yes	Source
Microsoft	Phi 3 Vision 128K Instruct	-	78.10%	inferred family alias from phi-3.5-vision-instruct (score=0.3495; benches=9)	Yes	Source
Microsoft	Phi 3.5 vision instruct	23 Aug 2024	78.10%	-	Yes	Source
DeepSeek	DeepSeek VL2 Small	13 Dec 2024	80%	-	Yes	Source
DeepSeek	DeepSeek VL2	13 Dec 2024	81.40%	-	Yes	Source
Microsoft	Phi 4 multimodal instruct	01 Feb 2025	82.30%	-	Yes	Source
Qwen	Qwen 2.5 Coder 3B	-	83.20%	inferred family alias from qwen2.5-omni-7b (score=0.3000; benches=45)	Yes	Source
Qwen	Qwen 2.5 Coder 7B	-	83.20%	inferred high-confidence family alias from qwen2.5-omni-7b (score=0.4700; benches=45)	Yes	Source
Qwen	Qwen 2.5 Math 7B	-	83.20%	inferred high-confidence family alias from qwen2.5-omni-7b (score=0.4767; benches=45)	Yes	Source
Qwen	Qwen 2.5 Math 7B PRM800K	-	83.20%	inferred family alias from qwen2.5-omni-7b (score=0.3696; benches=45)	Yes	Source
Qwen	Qwen 2.5 Math PRM 7B	-	83.20%	inferred family alias from qwen2.5-omni-7b (score=0.4092; benches=45)	Yes	Source
Qwen	Qwen 2.5 Omni 3B	-	83.20%	inferred high-confidence family alias from qwen2.5-omni-7b (score=0.4933; benches=45)	Yes	Source
Qwen	Qwen 2.5 Omni 7B	-	83.20%	-	Yes	Source
Qwen	Qwen 3 VL 4B Instruct	-	84.10%	-	Yes	-
Qwen	Qwen 3 4B Instruct 2507	-	84.90%	inferred family alias from qwen3-vl-4b-thinking (score=0.3462; benches=48)	Yes	-
Qwen	Qwen 3 4B SafeRL	-	84.90%	inferred family alias from qwen3-vl-4b-thinking (score=0.3850; benches=48)	Yes	-
Qwen	Qwen 3 4B Thinking 2507	-	84.90%	inferred family alias from qwen3-vl-4b-thinking (score=0.3462; benches=48)	Yes	-
Qwen	Qwen 3 8B	-	84.90%	inferred high-confidence family alias from qwen3-vl-8b-thinking (score=0.4600; benches=50)	Yes	-
Qwen	Qwen 3 Embedding 4B	-	84.90%	inferred family alias from qwen3-vl-4b-thinking (score=0.3850; benches=48)	Yes	-
Qwen	Qwen 3 Embedding 8B	-	84.90%	inferred family alias from qwen3-vl-8b-thinking (score=0.3850; benches=50)	Yes	-
Qwen	Qwen 3 Guard Gen 4B	-	84.90%	inferred family alias from qwen3-vl-4b-thinking (score=0.3400; benches=48)	Yes	-
Qwen	Qwen 3 Guard Gen 8B	-	84.90%	inferred family alias from qwen3-vl-8b-thinking (score=0.3400; benches=50)	Yes	-
Qwen	Qwen 3 Guard Stream 4B	-	84.90%	inferred family alias from qwen3-vl-4b-thinking (score=0.3371; benches=48)	Yes	-
Qwen	Qwen 3 Guard Stream 8B	-	84.90%	inferred family alias from qwen3-vl-8b-thinking (score=0.3371; benches=50)	Yes	-
Qwen	Qwen 3 Reranker 4B	-	84.90%	inferred family alias from qwen3-vl-4b-thinking (score=0.3850; benches=48)	Yes	-
Qwen	Qwen 3 Reranker 8B	-	84.90%	inferred family alias from qwen3-vl-8b-thinking (score=0.3850; benches=50)	Yes	-
Qwen	Qwen 3 VL 4B Thinking	-	84.90%	-	Yes	-
Qwen	Qwen 3 VL 8B Thinking	-	84.90%	-	Yes	-
Qwen	Qwen 3 VL Embedding 8B	-	84.90%	inferred high-confidence family alias from qwen3-vl-8b-thinking (score=0.5232; benches=50)	Yes	-
Qwen	Qwen 3 VL Reranker 8B	-	84.90%	inferred high-confidence family alias from qwen3-vl-8b-thinking (score=0.5275; benches=50)	Yes	-
Qwen	Qwen 3 VL 30B A3B Instruct	-	85%	-	Yes	-
Qwen	Qwen 3 VL 8B Instruct	-	85.70%	-	Yes	-
Qwen	Qwen 3 VL 30B A3B Thinking	-	86.90%	-	Yes	-
SpaceXAI	Grok 1.5V	12 Apr 2024	88.30%	-	Yes	Source
Qwen	Qwen 2.5 Math 72B	-	88.40%	inferred high-confidence family alias from qwen2.5-vl-72b (score=0.4700; benches=30)	Yes	Source
Qwen	Qwen 2.5 Math PRM 72B	-	88.40%	inferred family alias from qwen2.5-vl-72b (score=0.4050; benches=30)	Yes	Source
Qwen	Qwen 2.5 Math RM 72B	-	88.40%	inferred family alias from qwen2.5-vl-72b (score=0.4092; benches=30)	Yes	Source
Qwen	Qwen 2.5 VL 72B Instruct	26 Jan 2025	88.40%	-	Yes	Source
Qwen	Qwen 3 VL 32B Thinking	-	88.90%	-	Yes	-
Qwen	Qwen 3 VL 235B A22B Thinking	-	89.20%	-	Yes	-
Qwen	Qwen 3 VL 32B Instruct	-	89.50%	-	Yes	-
Qwen	Qwen 3 VL 235B A22B Instruct	-	89.70%	-	Yes	-
Meta	Llama 3.2 11B Vision Instruct	-	91.10%	inferred high-confidence family alias from llama-3.2-11b-instruct (score=0.5873; benches=11)	Yes	Source
Meta	Llama 3.2 1B Instruct	25 Sept 2024	91.10%	inferred family alias from llama-3.2-11b-instruct (score=0.4200; benches=11)	Yes	Source
Meta	Llama 3.2 90B Vision Instruct	-	92.30%	inferred high-confidence family alias from llama-3.2-90b-instruct (score=0.5873; benches=13)	Yes	Source
Qwen	Qwen 3.5 35B A3B	24 Feb 2026	92.60%	-	Yes	Source
Qwen	Qwen 3.5 27B	24 Feb 2026	92.90%	-	Yes	Source
Qwen	Qwen 3.5 Flash	23 Feb 2026	92.90%	inferred family alias from qwen3.5-27b (score=0.4147; benches=81)	Yes	Source
Mistral	Mistral Small 3.2	20 Jun 2025	92.91%	-	Yes	Source
Qwen	Qwen 3.5 122B A10B	24 Feb 2026	93.30%	-	Yes	Source
Mistral	Pixtral Large	18 Nov 2024	93.80%	-	Yes	Source
OpenAI	GPT 4o (2024-08-06)	06 Aug 2024	94.20%	-	Yes	Source
OpenAI	GPT 4o Audio (2025-06-03)	03 Jun 2025	94.20%	inferred modality/version alias from gpt-4o-2024-08-06	Yes	Source
OpenAI	GPT 4o Audio (2024-10-01)	01 Oct 2024	94.20%	inferred modality/version alias from gpt-4o-2024-08-06	Yes	Source
OpenAI	GPT 4o Audio (2024-12-17)	17 Dec 2024	94.20%	inferred modality/version alias from gpt-4o-2024-08-06	Yes	Source
OpenAI	GPT 4o Realtime Preview (2025-06-03)	03 Jun 2025	94.20%	inferred modality/version alias from gpt-4o-2024-08-06	Yes	Source
OpenAI	GPT 4o Realtime Preview (2024-10-01)	01 Oct 2024	94.20%	inferred modality/version alias from gpt-4o-2024-08-06	Yes	Source
OpenAI	GPT 4o Search Preview	11 Mar 2025	94.20%	inferred modality/version alias from gpt-4o-2024-08-06	Yes	Source
OpenAI	GPT 4o Transcribe	20 Mar 2025	94.20%	inferred modality/version alias from gpt-4o-2024-08-06	Yes	Source
OpenAI	GPT 4o Transcribe Diarize	15 Oct 2025	94.20%	inferred modality/version alias from gpt-4o-2024-08-06	Yes	Source
Qwen	Qwen 3.6 Plus	01 Apr 2026	94.40%	-	Yes	Source