Benchmark Leaderboard

GPQA

117 models with GPQA scores, ranked from highest to lowest.

Score Leaderboard

#	Model	Developer	Score
1	Google: Gemini 2.5 Pro	google	88.7
2	DeepSeek: DeepSeek V3.2 Speciale	deepseek	87.1
3	OpenAI: GPT-5.2-Codex	openai	86.0
4	OpenAI: GPT-5.1-Codex	openai	86.0
5	OpenAI: GPT-5	openai	85.4
6	MoonshotAI: Kimi K2 Thinking	moonshotai	83.8
7	OpenAI: GPT-5 Codex	openai	83.7
8	Anthropic: Claude 3.7 Sonnet (thinking)	anthropic	83.4
9	MiniMax: MiniMax M2.5	minimax	83.0
10	OpenAI: GPT-5 Mini	openai	82.8
11	OpenAI: o3	openai	82.7
12	Google: Gemini 2.5 Pro Preview 05-06	google	82.2
13	DeepSeek: R1	deepseek	81.3
14	OpenAI: GPT-5.1-Codex-Mini	openai	81.3
15	Google: Gemini 2.5 Flash	google	81.2
16	Baidu: ERNIE 4.5 300B A47B	baidu	81.1
17	xAI: Grok 3 Mini	x-ai	79.1
18	Qwen: Qwen3 235B A22B Thinking 2507	qwen	79.0
19	OpenAI: o4 Mini	openai	78.4
20	OpenAI: gpt-oss-120b	openai	78.2
21	MiniMax: MiniMax M2	minimax	77.7
22	Qwen: Qwen3 Max Thinking	qwen	77.6
23	OpenAI: o3 Mini High	openai	77.3
24	Qwen: Qwen3 VL 235B A22B Thinking	qwen	77.2
25	MoonshotAI: Kimi K2 0905	moonshotai	76.7
26	MoonshotAI: Kimi K2.5	moonshotai	76.6
27	Qwen: Qwen3 Max	qwen	76.4
28	Kwaipilot: KAT-Coder-Pro V1	kwaipilot	76.4
29	Prime Intellect: INTELLECT-3	prime-intellect	76.1
30	Qwen: Qwen3 Next 80B A3B Thinking	qwen	75.9
31	NVIDIA: Nemotron 3 Nano 30B A3B (free)	nvidia	75.7
32	Qwen: Qwen3 235B A22B Instruct 2507	qwen	75.3
33	DeepSeek: DeepSeek V3.2	deepseek	75.1
34	DeepSeek: DeepSeek V3.1 Terminus	deepseek	75.1
35	NVIDIA: Llama 3.3 Nemotron Super 49B V1.5	nvidia	74.8
36	OpenAI: o3 Mini	openai	74.8
37	OpenAI: o1	openai	74.7
38	Qwen: Qwen3 Next 80B A3B Instruct	qwen	73.8
39	Z.ai: GLM 4.5 Air	z-ai	73.3
40	Qwen: Qwen3 VL 32B Instruct	qwen	73.3
41	Nous: Hermes 3 405B Instruct	nousresearch	72.7
42	xAI: Grok Code Fast 1	x-ai	72.7
43	Qwen: Qwen3 VL 30B A3B Thinking	qwen	72.0
44	Z.ai: GLM 4.6V	z-ai	71.9
45	Qwen: Qwen3 VL 235B A22B Instruct	qwen	71.2
46	Qwen: Qwen3 30B A3B Thinking 2507	qwen	70.7
47	Qwen: Qwen3 VL 30B A3B Instruct	qwen	69.5
48	Upstage: Solar Pro 3	upstage	68.7
49	OpenAI: GPT-4.1 Mini	openai	68.7
50	MiniMax: MiniMax M1	minimax	68.2
51	Mistral Large	mistralai	68.0
52	OpenAI: GPT-5 Nano	openai	67.6
53	Anthropic: Claude Haiku 4.5	anthropic	67.2
54	Qwen: Qwen3 30B A3B	qwen	65.9
55	Xiaomi: MiMo-V2-Flash	xiaomi	65.6
56	Google: Gemini 2.5 Flash Lite Preview 09-2025	google	65.1
57	xAI: Grok 4 Fast	x-ai	63.7
58	xAI: Grok 4.1 Fast	x-ai	63.7
59	Google: Gemini 2.5 Flash Lite	google	62.5
60	Qwen: Qwen3 30B A3B Instruct 2507	qwen	62.0
61	DeepSeek: R1 Distill Qwen 32B	deepseek	61.5
62	Qwen: Qwen3 235B A22B	qwen	61.3
63	OpenAI: gpt-oss-20b	openai	61.1
64	Anthropic: Claude Sonnet 4.6	anthropic	59.9
65	AllenAI: Olmo 3.1 32B Think	allenai	59.1
66	Mistral: Mistral Medium 3.1	mistralai	58.8
67	Qwen: Qwen3 VL 8B Thinking	qwen	57.9
68	Perplexity: Sonar Pro	perplexity	57.8
69	Z.ai: GLM 4.5V	z-ai	57.3
70	NVIDIA: Nemotron Nano 12B 2 VL (free)	nvidia	57.2
71	NVIDIA: Nemotron Nano 9B V2	nvidia	57.0
72	NVIDIA: Nemotron Nano 9B V2 (free)	nvidia	57.0
73	Qwen: QwQ 32B	qwen	55.7
74	OpenAI: GPT-4o	openai	54.3
75	Qwen: Qwen3 32B	qwen	53.5
76	Google: Gemini 2.0 Flash Lite	google	53.5
77	Mistral: Devstral Small 1.1	mistralai	53.2
78	Cohere: Command A	cohere	52.7
79	OpenAI: GPT-4o (2024-05-13)	openai	52.6
80	Qwen: Qwen3 Coder 30B A3B Instruct	qwen	51.6
81	AllenAI: Olmo 3 7B Think	allenai	51.6
82	Meta: Llama 3.1 405B Instruct	meta-llama	51.5
83	Mistral: Pixtral Large 2411	mistralai	50.5
84	NVIDIA: Llama 3.1 Nemotron 70B Instruct	nvidia	49.8
85	Meta: Llama 3.3 70B Instruct	meta-llama	49.8
86	Mistral: Devstral Medium	mistralai	49.2
87	Qwen: Qwen2.5 VL 72B Instruct	qwen	49.1
88	Nous: Hermes 3 70B Instruct	nousresearch	49.1
89	Anthropic: Claude Opus 4.5	anthropic	48.9
90	Anthropic: Claude Opus 4.6	anthropic	48.9
91	Mistral Large 2407	mistralai	47.2
92	xAI: Grok 3 Beta	x-ai	47.1
93	Perplexity: Sonar	perplexity	47.1
94	Qwen: Qwen3 14B	qwen	47.0
95	Qwen: Qwen3 8B	qwen	45.2
96	NVIDIA: Nemotron Nano 12B 2 VL	nvidia	43.9
97	Qwen: Qwen3 VL 8B Instruct	qwen	42.7
98	OpenAI: GPT-4o-mini	openai	42.6
99	Mistral: Saba	mistralai	42.4
100	Qwen2.5 Coder 32B Instruct	qwen	41.7
101	Qwen: Qwen2.5 VL 32B Instruct	qwen	41.7
102	Qwen: Qwen-Turbo	qwen	41.0
103	DeepSeek: R1 Distill Llama 70B	deepseek	40.2
104	AllenAI: Olmo 3 7B Instruct	allenai	40.0
105	NVIDIA: Nemotron 3 Nano 30B A3B	nvidia	39.9
106	AI21: Jamba Large 1.7	ai21	39.0
107	LiquidAI: LFM2-8B-A1B	liquid	34.4
108	Qwen: Qwen2.5 Coder 7B Instruct	qwen	33.9
109	Cohere: Command R+ (08-2024)	cohere	32.3
110	OpenAI: GPT-3.5 Turbo	openai	29.7
111	Google: Gemma 3n 4B	google	29.6
112	Mistral: Mixtral 8x7B Instruct	mistralai	29.2
113	Meta: Llama 3.1 8B Instruct	meta-llama	25.9
114	Meta: Llama 3.2 3B Instruct	meta-llama	25.5
115	Meta: Llama 3.2 11B Vision Instruct	meta-llama	22.1
116	Meta: Llama 3.2 1B Instruct	meta-llama	19.6
117	Mistral: Mistral 7B Instruct v0.1	mistralai	17.7

Value Rankings

Best performance per dollar — score divided by avg price per million tokens.

#	Model	Score	Avg $/M	Value
1	LiquidAI: LFM2-8B-A1B	34.4	$0.01	2293.3
2	Google: Gemma 3n 4B	29.6	$0.03	986.7
3	Qwen: Qwen3 235B A22B Instruct 2507	75.3	$0.09	880.7
4	Meta: Llama 3.1 8B Instruct	25.9	$0.03	740.0
5	OpenAI: gpt-oss-20b	61.1	$0.09	718.8
6	OpenAI: gpt-oss-120b	78.2	$0.11	683.0
7	NVIDIA: Nemotron Nano 9B V2	57.0	$0.10	570.0
8	Qwen: Qwen2.5 Coder 7B Instruct	33.9	$0.06	565.0
9	Qwen: Qwen-Turbo	41.0	$0.08	504.6
10	Meta: Llama 3.2 11B Vision Instruct	22.1	$0.05	451.0
11	Qwen: Qwen3 30B A3B	65.9	$0.18	366.1
12	Qwen: Qwen3 30B A3B Thinking 2507	70.7	$0.20	361.6
13	Xiaomi: MiMo-V2-Flash	65.6	$0.19	345.3
14	Qwen: Qwen3 32B	53.5	$0.16	334.4
15	AllenAI: Olmo 3 7B Think	51.6	$0.16	322.5
16	NVIDIA: Nemotron 3 Nano 30B A3B	39.9	$0.12	319.2
17	Qwen: Qwen3 30B A3B Instruct 2507	62.0	$0.20	317.9
18	Qwen: Qwen3 14B	47.0	$0.15	313.3
19	Qwen: Qwen3 Coder 30B A3B Instruct	51.6	$0.17	303.5
20	OpenAI: GPT-5 Nano	67.6	$0.22	300.4
21	NVIDIA: Llama 3.3 Nemotron Super 49B V1.5	74.8	$0.25	299.2
22	Google: Gemini 2.0 Flash Lite	53.5	$0.19	285.3
23	Qwen: Qwen3 VL 32B Instruct	73.3	$0.26	281.9
24	AllenAI: Olmo 3 7B Instruct	40.0	$0.15	266.7
25	Mistral: Devstral Small 1.1	53.2	$0.20	266.0
26	Google: Gemini 2.5 Flash Lite Preview 09-2025	65.1	$0.25	260.4
27	Google: Gemini 2.5 Flash Lite	62.5	$0.25	250.0
28	Meta: Llama 3.3 70B Instruct	49.8	$0.21	237.1
29	DeepSeek: DeepSeek V3.2	75.1	$0.32	231.1
30	Qwen: Qwen3 235B A22B Thinking 2507	79.0	$0.35	222.5
31	Qwen: Qwen3 VL 30B A3B Instruct	69.5	$0.33	213.8
32	DeepSeek: R1 Distill Qwen 32B	61.5	$0.29	212.1
33	Qwen2.5 Coder 32B Instruct	41.7	$0.20	208.5
34	Qwen: QwQ 32B	55.7	$0.27	202.5
35	Qwen: Qwen3 8B	45.2	$0.22	200.9
36	xAI: Grok 3 Mini	79.1	$0.40	197.7
37	Upstage: Solar Pro 3	68.7	$0.38	183.2
38	xAI: Grok 4 Fast	63.7	$0.35	182.0
39	xAI: Grok 4.1 Fast	63.7	$0.35	182.0
40	AllenAI: Olmo 3.1 32B Think	59.1	$0.33	181.8
41	Meta: Llama 3.2 1B Instruct	19.6	$0.11	172.7
42	Nous: Hermes 3 70B Instruct	49.1	$0.30	163.7
43	DeepSeek: DeepSeek V3.1 Terminus	75.1	$0.50	150.2
44	Z.ai: GLM 4.5 Air	73.3	$0.49	149.6
45	Kwaipilot: KAT-Coder-Pro V1	76.4	$0.52	147.6
46	Qwen: Qwen3 VL 8B Instruct	42.7	$0.29	147.2
47	Qwen: Qwen3 VL 235B A22B Instruct	71.2	$0.54	131.9
48	Meta: Llama 3.2 3B Instruct	25.5	$0.20	130.4
49	Qwen: Qwen3 Next 80B A3B Instruct	73.8	$0.60	124.0
50	MiniMax: MiniMax M2	77.7	$0.63	123.8
51	Z.ai: GLM 4.6V	71.9	$0.60	119.8
52	Mistral: Mistral 7B Instruct v0.1	17.7	$0.15	118.0
53	Baidu: ERNIE 4.5 300B A47B	81.1	$0.69	117.5
54	Prime Intellect: INTELLECT-3	76.1	$0.65	117.1
55	OpenAI: GPT-4o-mini	42.6	$0.38	113.6
56	Qwen: Qwen3 Next 80B A3B Thinking	75.9	$0.67	112.4
57	MiniMax: MiniMax M2.5	83.0	$0.75	111.0
58	NVIDIA: Nemotron Nano 12B 2 VL	43.9	$0.40	109.8
59	DeepSeek: DeepSeek V3.2 Speciale	87.1	$0.80	108.9
60	Mistral: Saba	42.4	$0.40	106.0
61	Qwen: Qwen2.5 VL 32B Instruct	41.7	$0.40	104.3
62	xAI: Grok Code Fast 1	72.7	$0.85	85.5
63	Qwen: Qwen3 VL 8B Thinking	57.9	$0.74	78.1
64	OpenAI: GPT-5 Mini	82.8	$1.13	73.6
65	Nous: Hermes 3 405B Instruct	72.7	$1.00	72.7
66	OpenAI: GPT-5.1-Codex-Mini	81.3	$1.13	72.3
67	OpenAI: GPT-4.1 Mini	68.7	$1.00	68.7
68	MoonshotAI: Kimi K2 Thinking	83.8	$1.23	67.9
69	MoonshotAI: Kimi K2 0905	76.7	$1.20	63.9
70	Qwen: Qwen2.5 VL 72B Instruct	49.1	$0.80	61.4
71	Google: Gemini 2.5 Flash	81.2	$1.40	58.0
72	MoonshotAI: Kimi K2.5	76.6	$1.33	57.8
73	Mistral: Mixtral 8x7B Instruct	29.2	$0.54	54.1
74	Qwen: Qwen3 235B A22B	61.3	$1.14	53.9
75	DeepSeek: R1 Distill Llama 70B	40.2	$0.75	53.6
76	MiniMax: MiniMax M1	68.2	$1.30	52.5
77	DeepSeek: R1	81.3	$1.60	50.8
78	Mistral: Mistral Medium 3.1	58.8	$1.20	49.0
79	Z.ai: GLM 4.5V	57.3	$1.20	47.8
80	Perplexity: Sonar	47.1	$1.00	47.1
81	NVIDIA: Llama 3.1 Nemotron 70B Instruct	49.8	$1.20	41.5
82	Mistral: Devstral Medium	49.2	$1.20	41.0
83	Qwen: Qwen3 Max Thinking	77.6	$2.34	33.2
84	OpenAI: GPT-3.5 Turbo	29.7	$1.00	29.7
85	OpenAI: o4 Mini	78.4	$2.75	28.5
86	OpenAI: o3 Mini High	77.3	$2.75	28.1
87	OpenAI: o3 Mini	74.8	$2.75	27.2
88	Anthropic: Claude Haiku 4.5	67.2	$3.00	22.4
89	Qwen: Qwen3 Max	76.4	$3.60	21.2
90	Mistral Large	68.0	$4.00	17.0
91	OpenAI: o3	82.7	$5.00	16.5
92	Google: Gemini 2.5 Pro	88.7	$5.63	15.8
93	OpenAI: GPT-5.1-Codex	86.0	$5.63	15.3
94	OpenAI: GPT-5	85.4	$5.63	15.2
95	OpenAI: GPT-5 Codex	83.7	$5.63	14.9
96	Google: Gemini 2.5 Pro Preview 05-06	82.2	$5.63	14.6
97	Meta: Llama 3.1 405B Instruct	51.5	$4.00	12.9
98	Mistral: Pixtral Large 2411	50.5	$4.00	12.6
99	Mistral Large 2407	47.2	$4.00	11.8
100	OpenAI: GPT-5.2-Codex	86.0	$7.88	10.9
101	Anthropic: Claude 3.7 Sonnet (thinking)	83.4	$9.00	9.3
102	OpenAI: GPT-4o	54.3	$6.25	8.7
103	Cohere: Command A	52.7	$6.25	8.4
104	AI21: Jamba Large 1.7	39.0	$5.00	7.8
105	Anthropic: Claude Sonnet 4.6	59.9	$9.00	6.7
106	Perplexity: Sonar Pro	57.8	$9.00	6.4
107	OpenAI: GPT-4o (2024-05-13)	52.6	$10.00	5.3
108	xAI: Grok 3 Beta	47.1	$9.00	5.2
109	Cohere: Command R+ (08-2024)	32.3	$6.25	5.2
110	Anthropic: Claude Opus 4.5	48.9	$15.00	3.3
111	Anthropic: Claude Opus 4.6	48.9	$15.00	3.3
112	OpenAI: o1	74.7	$37.50	2.0