Knowledge & reasoning

Vals GPQA Diamond leaderboard

Name: benchgap: LLM benchmark scores, measured and estimated
Creator: benchgap

As of 2026-10-07, the highest measured score on Vals GPQA Diamond is 95.5% by Gemini 3.1 Pro. 185 more models have estimated scores, calibrated from the benchmarks they were measured on.

Measured scores: benchlm.ai.

#	Model	Score	Source
1	Claude Sonnet 5.5	98.2%	estimated ± 1.5 pp, low confidence
2	Claude Opus 5.5	98.2%	estimated ± 1.5 pp, low confidence
3	Ling 3.1 Flash	98.2%	estimated ± 1.5 pp, low confidence
4	GPT-6.1 Sol	98.2%	estimated ± 1.5 pp, low confidence
5	Gemini 4 Argon	96.4%	estimated ± 2.2 pp, low confidence
6	GPT-6 Luna	96.2%	estimated ± 1.5 pp, low confidence
7	GPT-6 Sol	96.2%	estimated ± 1.5 pp, low confidence
8	GPT-6 Astra	96.1%	estimated ± 1.1 pp, high confidence
9	Gemini 3.1 Pro	95.5%	measured
10	GPT-5.6 Sol	95.2%	measured
11	Muse Spark 1.3	95.2%	estimated ± 7.4 pp, medium confidence
12	Sakana Fugu	94.9%	estimated ± 2.0 pp, medium confidence
13	Sakana Fugu-Ultra	94.9%	estimated ± 2.0 pp, medium confidence
14	Grok 4.6	94.7%	measured
15	Claude Mythos 5	94.4%	estimated ± 3.0 pp, high confidence
16	Gemini 3.8 Flash	94.4%	measured
17	Gemini 3.7 Flash	93.9%	measured
18	Qwen3.8 Max Preview	93.8%	estimated ± 7.4 pp, medium confidence
19	Qwen3.8 Max	93.7%	measured
20	GPT-5.5 Pro	93.7%	estimated ± 2.2 pp, high confidence
21	GPT-5.4 Pro	93.6%	estimated ± 2.2 pp, high confidence
22	Claude Fable 5.1	93.4%	measured
23	Claude Opus 5	93.4%	measured
24	Gemini 3.6 Flash	93.4%	measured
25	dots3-note Preview	93.4%	estimated ± 2.2 pp, high confidence
26	Claude Fable 5	93.2%	measured
27	GPT-5.5	93.2%	measured
28	Grok 4.5	92.9%	measured
29	Kimi K3	92.9%	measured
30	Step 5 Preview	92.8%	estimated ± 2.0 pp, high confidence
31	Gemini 3.5 Flash	92.7%	measured
32	MiniMax M3	92.7%	measured
33	Claude Opus 4.8	92.4%	measured
34	DeepSeek V4 Pro 0813	92.4%	measured
35	Pareto 26.9	92.3%	estimated ± 3.0 pp, high confidence
36	GPT-5.4	92.2%	estimated ± 1.1 pp, high confidence
37	Ornith-1.5-397B	92.1%	estimated ± 2.0 pp, high confidence
38	Claude Opus 4.6 (Adaptive)	92.1%	estimated ± 2.2 pp, high confidence
39	Pareto 26.10 Preview	91.7%	estimated ± 2.0 pp, high confidence
40	GPT-5.6 Luna	91.7%	measured
41	Claude Opus 4.7 (Adaptive)	91.7%	estimated ± 1.1 pp, high confidence
42	Hy4 preview	91.6%	estimated ± 2.0 pp, high confidence
43	Claude Haiku 5.5	91.5%	estimated ± 3.0 pp, high confidence
44	MiMo-V2.6-Pro	91.4%	estimated ± 7.5 pp, medium confidence
45	GPT-5.3 Codex	91.4%	estimated ± 7.4 pp, medium confidence
46	Grok 4.3	91.4%	measured
47	Muse Spark 1.1	91.2%	measured
48	Mistral Large 4	91.2%	estimated ± 7.5 pp, medium confidence
49	MiMo-V2.6-Flash	91.1%	estimated ± 7.5 pp, medium confidence
50	Grok 4.7	91.1%	estimated ± 1.5 pp, medium confidence
51	Qwen3.8-Flash-Next	91.0%	estimated ± 2.0 pp, high confidence
52	GPT-5.6 Terra	90.9%	measured
53	Qwen3.8-Omni-Flash	90.3%	estimated ± 2.0 pp, high confidence
54	Claude Opus 4.7	90.2%	measured
55	Qwen3.7 Max	90.2%	measured
56	DeepSeek V4.1 Flash	90.2%	estimated ± 2.0 pp, high confidence
57	GPT-5.2	90.0%	estimated ± 2.2 pp, high confidence
58	Muse Spark 1.2	90.0%	estimated ± 3.8 pp, high confidence
59	DeepSeek V4 Flash 0731	89.9%	measured
60	Beam	89.8%	estimated ± 2.0 pp, high confidence
61	Muse Spark	89.6%	measured
62	Qwen3.7 Plus	89.6%	estimated ± 2.0 pp, high confidence
63	Interfaze Beta	89.1%	estimated ± 2.0 pp, high confidence
64	Kimi K2.6	89.1%	measured
65	Claude Sonnet 5	88.9%	measured
66	Qwen3.8-27B	88.9%	measured
67	Gemini 3 Pro Deep Think	88.8%	estimated ± 2.2 pp, high confidence
68	Grok 4.20	88.6%	measured
69	GPT-5.2-Codex	88.5%	estimated ± 7.4 pp, medium confidence
70	Claude Opus 4.6	88.4%	estimated ± 2.0 pp, high confidence
71	Ornith-1.5-35B-A3B	88.4%	estimated ± 2.0 pp, high confidence
72	Solar Pro 4	88.2%	estimated ± 2.0 pp, high confidence
73	Hy3	88.2%	estimated ± 7.4 pp, medium confidence
74	GLM-5.3	88.1%	measured
75	Kimi K2.7 Code	88.0%	estimated ± 7.4 pp, medium confidence
76	Gemini 3 Flash	87.9%	measured
77	Qwen3.6 Plus	87.4%	measured
78	Claude Opus 4.5 Thinking	87.3%	estimated ± 2.2 pp, high confidence
79	Inkling	87.1%	measured
80	Kimi K2.5	86.8%	estimated ± 2.0 pp, high confidence
81	MiniMax M2.7	86.6%	measured
82	Qwen 3.6 Max (preview)	86.6%	estimated ± 7.4 pp, medium confidence
83	GLM-5.3-Flash	86.4%	measured
84	Hy3 Preview	86.4%	estimated ± 2.0 pp, high confidence
85	Nemotron 3 Ultra	86.1%	measured
86	Gemini 3 Pro	85.9%	estimated ± 2.2 pp, high confidence
87	Claude Sonnet 4.6	85.6%	measured
88	GLM-5.2	85.6%	measured
89	Ornith-1.5-9B	85.6%	estimated ± 2.0 pp, high confidence
90	Solar Open 2	85.4%	estimated ± 2.0 pp, high confidence
91	GPT-5.1	85.3%	estimated ± 2.3 pp, high confidence
92	GLM-5	85.1%	estimated ± 2.0 pp, high confidence
93	Ternary Bonsai 2 27B	84.9%	estimated ± 2.0 pp, high confidence
94	Ling 3.0 Flash	84.8%	measured
95	A.X K2	84.7%	estimated ± 2.0 pp, high confidence
96	Grok 4	84.7%	estimated ± 7.4 pp, medium confidence
97	Qwen3.5 397B	84.6%	estimated ± 4.8 pp, high confidence
98	GLM-5.1	84.5%	measured
99	Gemini 3.5 Flash-Lite	83.8%	measured
100	Inkling-Small	83.6%	measured
101	Qwen3.6-27B	83.5%	estimated ± 4.8 pp, high confidence
102	MiMo-V2-Pro	83.5%	estimated ± 7.4 pp, medium confidence
103	MAI-Thinking-1	83.3%	estimated ± 2.0 pp, medium confidence
104	Kimi K2.5 (Reasoning)	83.1%	estimated ± 4.8 pp, high confidence
105	GPT-5.4 mini	83.1%	measured
106	Ling 3.0 Flash FP8	83.1%	estimated ± 2.0 pp, medium confidence
107	Qwen3.5 Flash	82.8%	measured
108	MiMo-V2.5-Pro	82.6%	measured
109	Apodex 1.1	82.4%	estimated ± 7.4 pp, medium confidence
110	Apodex 1.1 Mini	82.4%	estimated ± 7.4 pp, medium confidence
111	Ling 3.0 Flash VL	82.1%	estimated ± 7.4 pp, medium confidence
112	Claude Opus 4.5	82.1%	estimated ± 4.8 pp, high confidence
113	Qwen3.5 397B (Reasoning)	81.9%	estimated ± 7.4 pp, medium confidence
114	GPT-5.1-Codex	81.8%	estimated ± 7.4 pp, medium confidence
115	GPT-5.1-Codex-Max	81.8%	estimated ± 7.4 pp, medium confidence
116	MiMo-V2.5	81.6%	measured
117	Qwen3.5-122B-A10B	81.4%	estimated ± 4.8 pp, high confidence
118	K-EXAONE 2.0	81.2%	estimated ± 2.0 pp, medium confidence
119	Gemini 3.1 Flash-Lite	81.1%	measured
120	GPT-5 (high)	80.8%	estimated ± 7.4 pp, medium confidence
121	Claude Sonnet 4.5 Thinking	80.7%	estimated ± 2.2 pp, high confidence
122	Claude Sonnet 4.5	80.7%	estimated ± 2.2 pp, high confidence
123	Grok 4.1 Fast (Reasoning)	80.6%	estimated ± 7.4 pp, medium confidence
124	Qwen3.6-35B-A3B	80.3%	estimated ± 4.8 pp, high confidence
125	GLM-4.7	80.1%	measured
126	GLM-5-Turbo	79.6%	estimated ± 7.4 pp, medium confidence
127	Grok 4 Fast (Reasoning)	79.6%	estimated ± 7.4 pp, medium confidence
128	Qwen3.5-27B	79.5%	estimated ± 4.8 pp, high confidence
129	o3-pro	79.3%	estimated ± 7.4 pp, medium confidence
130	GPT-5 (medium)	78.8%	estimated ± 7.4 pp, medium confidence
131	Mercury 2.5	78.0%	estimated ± 2.0 pp, medium confidence
132	Gemma 4 12B	77.7%	estimated ± 2.0 pp, medium confidence
133	Muse Glimmer 30B	77.7%	estimated ± 7.4 pp, medium confidence
134	GPT-5.4 nano	77.5%	measured
135	Qwen3.5-35B-A3B	77.3%	estimated ± 4.8 pp, high confidence
136	Gemma 4 31B	76.9%	estimated ± 3.0 pp, medium confidence
137	MiMo-V2-Omni	76.6%	estimated ± 7.4 pp, medium confidence
138	Claude 4.1 Opus	76.5%	estimated ± 7.5 pp, medium confidence
139	o3	76.4%	estimated ± 7.4 pp, medium confidence
140	Gemini 2.5 Pro	75.3%	estimated ± 4.8 pp, high confidence
141	Trinity-Large-Thinking	75.2%	estimated ± 2.0 pp, medium confidence
142	GLM-4.6	74.5%	measured
143	Nemotron 3.5 Lightning 30B A3B NVFP4	74.4%	estimated ± 2.0 pp, medium confidence
144	DeepSeek-R1	74.2%	estimated ± 7.4 pp, medium confidence
145	Claude 4.1 Opus Thinking	73.6%	estimated ± 7.4 pp, medium confidence
146	GLM-5V-Turbo	73.6%	estimated ± 7.4 pp, medium confidence
147	Step 3.7 Flash	73.6%	estimated ± 7.4 pp, medium confidence
148	Nemotron 3 Super 100B	72.3%	estimated ± 7.4 pp, medium confidence
149	Claude Haiku 4.5	72.2%	measured
150	GLM-4.5	72.2%	measured
151	Nemotron 3 Nano Omni 30B A3B	71.0%	estimated ± 2.0 pp, medium confidence
152	K-Exaone	69.7%	estimated ± 7.4 pp, medium confidence
153	ZAYA1-8B	69.7%	estimated ± 2.0 pp, medium confidence
154	GPT-OSS 120B	69.6%	estimated ± 7.4 pp, medium confidence
155	DeepSeek V3.1 (Reasoning)	69.2%	estimated ± 7.4 pp, medium confidence
156	MiniCPM5-2B	68.9%	estimated ± 2.0 pp, medium confidence
157	o1-pro	68.8%	estimated ± 4.8 pp, medium confidence
158	LongCat-Flash-Lite-Sparse	68.2%	estimated ± 2.0 pp, medium confidence
159	Mistral Small 4	67.7%	estimated ± 7.4 pp, medium confidence
160	Mistral Small 4 (Reasoning)	67.7%	estimated ± 7.4 pp, medium confidence
161	Kimi K2	67.3%	estimated ± 7.4 pp, medium confidence
162	o1-preview	67.1%	estimated ± 7.4 pp, medium confidence
163	Qwen3 Max	67.0%	estimated ± 7.4 pp, medium confidence
164	Command A+	66.6%	estimated ± 7.4 pp, medium confidence
165	Qwen3 235B 2507	66.5%	estimated ± 4.8 pp, medium confidence
166	o3-mini	66.1%	estimated ± 4.8 pp, medium confidence
167	Nemotron 3 Nano 30B	66.0%	estimated ± 7.4 pp, medium confidence
168	North Mini Code	66.0%	estimated ± 7.4 pp, medium confidence
169	DeepSeek V3.2	65.2%	estimated ± 7.4 pp, medium confidence
170	o1	63.8%	estimated ± 4.8 pp, medium confidence
171	Sarvam 105B	63.4%	estimated ± 7.4 pp, medium confidence
172	DeepSeek V3.1	63.0%	estimated ± 7.4 pp, medium confidence
173	Ling 3.0 Tiny	62.9%	estimated ± 7.4 pp, medium confidence
174	GLM-4.5-Air	62.7%	estimated ± 7.4 pp, medium confidence
175	Quasar 438B	62.6%	estimated ± 7.4 pp, medium confidence
176	Nemotron Ultra 253B	62.0%	estimated ± 7.4 pp, medium confidence
177	Grok Code Fast 1	61.9%	estimated ± 7.4 pp, medium confidence
178	Trinity-Large-Preview	61.8%	estimated ± 2.0 pp, medium confidence
179	Qwen3-Omni-30B-A3B-Thinking	61.8%	estimated ± 7.4 pp, medium confidence
180	Solar Pro 3	61.5%	estimated ± 7.4 pp, medium confidence
181	MiMo-V2-Flash	59.3%	measured
182	Gemma 4 26B A4B	57.2%	estimated ± 3.0 pp, medium confidence
183	GPT-OSS 20B	56.8%	estimated ± 7.4 pp, medium confidence
184	Claude 4 Sonnet	56.2%	estimated ± 7.4 pp, medium confidence
185	Gemini 2.5 Flash	56.2%	estimated ± 7.4 pp, medium confidence
186	Mellum2-12B-A2.5B-Thinking	56.0%	estimated ± 2.0 pp, medium confidence
187	Mistral Large 3	55.8%	estimated ± 7.4 pp, medium confidence
188	ZAYA1-74B-Preview	55.7%	estimated ± 2.0 pp, medium confidence
189	Laguna XS.2	55.1%	measured
190	Llama 4 Maverick	54.7%	estimated ± 7.4 pp, medium confidence
191	DeepSeek V3 0324	52.8%	estimated ± 7.4 pp, medium confidence
192	Granite 4.2 30B	51.0%	estimated ± 4.8 pp, medium confidence
193	GPT-4.1	50.8%	estimated ± 4.8 pp, medium confidence
194	Grok 4.1 Fast	50.7%	estimated ± 7.4 pp, medium confidence
195	Sarvam 30B	50.2%	estimated ± 7.4 pp, medium confidence
196	Celeris-1	50.0%	estimated ± 7.4 pp, low confidence
197	Exaone 4.0 32B	49.6%	estimated ± 7.4 pp, low confidence
198	Qwen3-Omni-30B-A3B-Instruct	48.7%	estimated ± 7.4 pp, low confidence
199	DeepSeek R1 Distill Qwen 32B	48.1%	estimated ± 7.4 pp, low confidence
200	GPT-4.1 mini	48.1%	estimated ± 4.8 pp, medium confidence
201	Granite 4.2 8B	48.1%	estimated ± 4.8 pp, medium confidence
202	Gemini 1.5 Pro	45.3%	estimated ± 7.4 pp, low confidence
203	Llama 4 Scout	45.0%	estimated ± 7.4 pp, low confidence
204	Mistral Medium 3	44.1%	estimated ± 7.4 pp, low confidence
205	Phi-4	43.7%	estimated ± 7.4 pp, low confidence
206	LLaDA2.2-mini	42.4%	estimated ± 2.0 pp, medium confidence
207	Claude 3.5 Sonnet	42.3%	estimated ± 4.8 pp, medium confidence
208	Solar Pro 2	42.3%	estimated ± 7.4 pp, low confidence
209	LFM2.5-2.6B	41.9%	estimated ± 7.4 pp, low confidence
210	DeepSeek V3	41.9%	estimated ± 4.8 pp, medium confidence
211	Ling 2.6 Flash	41.8%	estimated ± 4.8 pp, medium confidence
212	Soofi S 30B-A3B	41.4%	estimated ± 2.0 pp, medium confidence
213	Gemma 4 E4B	41.3%	estimated ± 4.8 pp, medium confidence
214	GPT-4o	40.4%	estimated ± 7.4 pp, low confidence
215	Mellum2-12B-A2.5B-Instruct	38.8%	estimated ± 2.0 pp, medium confidence
216	Llama 3.1 405B	37.6%	estimated ± 7.4 pp, low confidence
217	LFM2.5-8B-A1B	37.4%	estimated ± 7.4 pp, low confidence
218	Granite 4.2 3B	37.0%	estimated ± 4.8 pp, medium confidence
219	Nova Pro	36.0%	estimated ± 7.4 pp, low confidence
220	Ultravox v0.6 Llama 3.3 70B	35.9%	estimated ± 7.4 pp, low confidence
221	Claude 3 Opus	35.1%	estimated ± 7.4 pp, low confidence
222	Mistral Medium 3.5 128B	34.8%	measured
223	Mistral Large 2	34.8%	estimated ± 7.4 pp, low confidence
224	GPT-4.1 nano	32.1%	estimated ± 4.8 pp, medium confidence
225	Gemma 3 27B	29.5%	estimated ± 7.4 pp, low confidence
226	GPT-4o mini	29.3%	estimated ± 7.4 pp, low confidence
227	Exaone 4.0 1.2B	29.2%	estimated ± 7.4 pp, low confidence
228	Qwen2.5 Coder 32B Instruct	28.6%	estimated ± 7.4 pp, low confidence
229	Laguna M.1	27.0%	measured
230	Gemma 4 E2B	25.1%	estimated ± 4.8 pp, medium confidence
231	Claude 3 Haiku	24.9%	estimated ± 7.4 pp, low confidence
232	MiniCPM5-1B	23.8%	estimated ± 2.0 pp, medium confidence
233	LFM2.5-230M	22.9%	estimated ± 2.0 pp, medium confidence
234	Phi-4 Multimodal Instruct	20.3%	estimated ± 7.4 pp, low confidence
235	LFM2.5-VL-1.6B-Extract	18.3%	estimated ± 7.4 pp, low confidence
236	Gemini 1.0 Pro	17.4%	estimated ± 7.4 pp, low confidence
237	Granite-4.0-H-1B	16.4%	estimated ± 7.4 pp, low confidence
238	Granite-4.0-350M	16.3%	estimated ± 7.4 pp, low confidence
239	Granite-4.0-H-350M	16.0%	estimated ± 7.4 pp, low confidence
240	LFM2.5-VL-450M	9.6%	estimated ± 4.8 pp, medium confidence
241	GPT-4 Turbo	3.0%	estimated ± 7.5 pp, low confidence

Vals GPQA Diamond leaderboard

All leaderboards

Agentic · terminal

Agentic · tools

Coding

Math

Knowledge & reasoning

Instruction following

Multilingual

Vision & documents

Long context

General