SuperGPQA

AI benchmark results and model performance

Type: percentage

Recorded Results

Average Score

55.06%

Score Range

21.30% - 71.60%

Leading Model (lowest score)

21.30% - Qwen 3.5 0.8B

Scores Over Time

Individual benchmark scores plotted by date.

Models Using This Benchmark

Organisation	Model	Reported	Top Score	Info	Self Reported	Source
Qwen	Qwen 3.5 0.8B	02 Mar 2026	21.30%	-	Yes	Source
Qwen	Qwen 3.5 2B	02 Mar 2026	37.50%	-	Yes	Source
Qwen	Qwen 3 VL 4B Instruct	-	40.30%	-	Yes	-
Qwen	Qwen 3 235B A22B	-	44.06%	-	Yes	Source
Qwen	Qwen 3 VL 8B Instruct	-	44.50%	-	Yes	-
Moonshot	Kimi K2 (2025-07-11)	11 Jul 2025	44.70%	EM	Yes	Source
Qwen	Qwen 3 4B Instruct 2507	-	46.80%	inferred family alias from qwen3-vl-4b-thinking (score=0.3462; benches=48)	Yes	-
Qwen	Qwen 3 4B SafeRL	-	46.80%	inferred family alias from qwen3-vl-4b-thinking (score=0.3850; benches=48)	Yes	-
Qwen	Qwen 3 4B Thinking 2507	-	46.80%	inferred family alias from qwen3-vl-4b-thinking (score=0.3462; benches=48)	Yes	-
Qwen	Qwen 3 Embedding 4B	-	46.80%	inferred family alias from qwen3-vl-4b-thinking (score=0.3850; benches=48)	Yes	-
Qwen	Qwen 3 Guard Gen 4B	-	46.80%	inferred family alias from qwen3-vl-4b-thinking (score=0.3400; benches=48)	Yes	-
Qwen	Qwen 3 Guard Stream 4B	-	46.80%	inferred family alias from qwen3-vl-4b-thinking (score=0.3371; benches=48)	Yes	-
Qwen	Qwen 3 Reranker 4B	-	46.80%	inferred family alias from qwen3-vl-4b-thinking (score=0.3850; benches=48)	Yes	-
Qwen	Qwen 3 VL 4B Thinking	-	46.80%	-	Yes	-
Qwen	Qwen 3 8B	-	51.20%	inferred high-confidence family alias from qwen3-vl-8b-thinking (score=0.4600; benches=50)	Yes	-
Qwen	Qwen 3 Guard Stream 8B	-	51.20%	inferred family alias from qwen3-vl-8b-thinking (score=0.3371; benches=50)	Yes	-
Qwen	Qwen 3 VL 8B Thinking	-	51.20%	-	Yes	-
Qwen	Qwen 3 VL Embedding 8B	-	51.20%	inferred high-confidence family alias from qwen3-vl-8b-thinking (score=0.5232; benches=50)	Yes	-
Qwen	Qwen 3 VL Reranker 8B	-	51.20%	inferred high-confidence family alias from qwen3-vl-8b-thinking (score=0.5275; benches=50)	Yes	-
Qwen	Qwen 3.5 4B	02 Mar 2026	52.90%	-	Yes	Source
Qwen	Qwen 3 VL 30B A3B Instruct	-	53.10%	-	Yes	-
Qwen	Qwen 3 VL 32B Instruct	-	54.60%	-	Yes	-
Qwen	Qwen 3 VL 30B A3B Thinking	-	56.40%	-	Yes	-
Moonshot	Kimi K2 (2025-09-05)	05 Sept 2025	57.20%	Acc	Yes	Source
Qwen	Qwen 3.5 9B	02 Mar 2026	58.20%	-	Yes	Source
Qwen	Qwen 3 Next 80B A3B Instruct	-	58.80%	-	Yes	Source
Qwen	Qwen 3 VL 32B Thinking	-	59%	-	Yes	-
Qwen	Qwen 3 VL 235B A22B Instruct	-	60.40%	-	Yes	-
Qwen	Qwen 3 Next 80B A3B Thinking	-	60.80%	-	Yes	Source
ByteDance	Seed 2.0 Mini	14 Feb 2026	61.60%	Seed2 official benchmark table \| SuperGPQA	Yes	Source
Qwen	Qwen 3 A235 A22B Instruct 2507	-	62.60%	-	Yes	Source
Qwen	Qwen 3.5 35B A3B	24 Feb 2026	63.40%	-	Yes	Source
Qwen	Qwen 3 VL 235B A22B Thinking	-	64.30%	-	Yes	-
Qwen	Qwen 3 235B A22B Thinking 2507	-	64.90%	-	Yes	Source
Qwen	Qwen 3 14B	-	65.10%	inferred family alias from qwen3-max (score=0.3333; benches=6)	Yes	Source
Qwen	Qwen 3 Max Thinking	26 Jan 2026	65.10%	inferred alias from qwen3-max	Yes	Source
Qwen	Qwen 3 TTS (2025-11-27)	-	65.10%	inferred family alias from qwen3-max (score=0.3833; benches=6)	Yes	Source
Qwen	Qwen 3.5 27B	24 Feb 2026	65.60%	-	Yes	Source
Qwen	Qwen 3.5 Flash	23 Feb 2026	65.60%	inferred family alias from qwen3.5-27b (score=0.4147; benches=81)	Yes	Source
Qwen	Qwen 3.5 122B A10B	24 Feb 2026	67.10%	-	Yes	Source
ByteDance	Seed 2.0 Lite	14 Feb 2026	67.50%	Seed2 official benchmark table \| SuperGPQA	Yes	Source
ByteDance	Seed 2.0 Pro	14 Feb 2026	68.70%	Seed2 official benchmark table \| SuperGPQA	Yes	Source
Qwen	Qwen 3.5 397B A17B	16 Feb 2026	70.40%	-	Yes	Source
Qwen	Qwen 3.6 Plus	01 Apr 2026	71.60%	-	Yes	Source