Agentic · tools

GDPval-AA leaderboard

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

As of 2026-10-07, the highest measured score on GDPval-AA is 68.3% by Claude Opus 5.5. 116 more models have estimated scores, calibrated from the benchmarks they were measured on.

Measured scores: benchlm.ai.

#	Model	Score	Source
1	BTL-3	100.0%	estimated ± 5.8 pp, low confidence
2	Claude Opus 5.5	68.3%	measured
3	Claude Sonnet 5.5	67.0%	measured
4	GLM-5.3-Flash	63.2%	estimated ± 4.6 pp, high confidence
5	Claude Fable 5.1	62.9%	measured
6	Claude Opus 5	61.2%	measured
7	Grok 4.7	60.8%	measured
8	MiMo-V2.6-Pro	59.3%	measured
9	Muse Spark 1.3	59.2%	measured
10	Qwen3.8 Max Preview	58.6%	measured
11	GLM-5.3	57.6%	measured
12	Qwen3.8-Flash-Next	56.6%	measured
13	Gemini 4 Argon	56.3%	measured
14	Grok 4.6	56.1%	measured
15	Ling 3.1 Flash	56.1%	measured
16	Claude Fable 5	55.6%	measured
17	GPT-5.6 Sol	55.6%	measured
18	MiMo-V2.6-Flash	55.5%	measured
19	DeepSeek V4.1 Flash	55.0%	measured
20	DeepSeek V4 Pro 0813	54.5%	measured
21	Step 5 Preview	54.3%	measured
22	GPT-6.1 Sol	53.8%	measured
23	Fugu Cyber	53.8%	estimated ± 5.1 pp, medium confidence
24	Gemini 3.8 Flash Cyber	53.2%	estimated ± 5.1 pp, medium confidence
25	GPT-5.5 Pro	52.3%	estimated ± 8.4 pp, medium confidence
26	GPT-6 Astra	52.1%	measured
27	Kimi K3	51.8%	measured
28	GPT-5.4 Pro	51.1%	estimated ± 8.4 pp, medium confidence
29	Claude Mythos 5	51.1%	estimated ± 5.1 pp, medium confidence
30	Gemini 3.5 Flash Cyber	50.5%	estimated ± 5.1 pp, medium confidence
31	GPT-6 Sol	50.5%	measured
32	Claude Mythos Preview	50.5%	estimated ± 5.1 pp, medium confidence
33	Muse Spark 1.2	48.9%	measured
34	Holo3-35B-A3B	48.8%	estimated ± 8.8 pp, medium confidence
35	Claude Sonnet 5	48.3%	measured
36	GPT-5.6 Luna	48.2%	measured
37	Claude Opus 4.8	47.8%	measured
38	GPT-5.6 Terra	47.7%	measured
39	DeepSeek V4 Flash 0731	46.9%	measured
40	GPT-6 Luna	46.9%	measured
41	Gemini 3.8 Flash	46.8%	measured
42	Mistral Large 4	46.2%	measured
43	Qwen3.8-27B	46.2%	measured
44	Gemini 3.7 Flash	44.6%	measured
45	Grok 4.5	44.5%	measured
46	Holo3-122B-A10B	43.9%	estimated ± 8.8 pp, medium confidence
47	GLM-5.2	43.7%	measured
48	Claude Opus 4.7	43.3%	estimated ± 7.5 pp, medium confidence
49	Claude Opus 4.7 (Adaptive)	42.8%	measured
50	GPT-5.5	42.7%	measured
51	Gemini 3.5 Flash	42.2%	measured
52	Claude Haiku 5.5	41.8%	estimated ± 6.9 pp, medium confidence
53	Ling 3.0 Flash Fin	39.5%	estimated ± 8.0 pp, low confidence
54	Gemini 3.6 Flash	39.3%	measured
55	Claude Opus 4.6 (Adaptive)	38.7%	estimated ± 11.2 pp, low confidence
56	UI-Mate-27B	38.1%	estimated ± 8.8 pp, medium confidence
57	GPT-5.4	37.4%	measured
58	MiniMax M3	37.3%	measured
59	GPT-5.2	37.3%	estimated ± 6.2 pp, medium confidence
60	GPT-5.3 Codex	36.8%	estimated ± 6.2 pp, medium confidence
61	Claude Opus 4.6	35.8%	estimated ± 5.1 pp, medium confidence
62	Hy3 Preview	35.8%	measured
63	Muse Spark 1.1	35.7%	measured
64	Beam	35.5%	estimated ± 8.4 pp, medium confidence
65	Agents-A1	35.4%	estimated ± 8.1 pp, medium confidence
66	Apodex 1.1	34.8%	measured
67	Apodex 1.1 Mini	34.8%	measured
68	Claude Sonnet 4.6	34.8%	estimated ± 5.1 pp, medium confidence
69	Quasar 438B	34.2%	measured
70	Ling 3.0 Flash VL	33.2%	measured
71	K-EXAONE 2.0	33.1%	estimated ± 9.4 pp, low confidence
72	Nemotron 3 Ultra	33.1%	measured
73	Claude Sonnet 4.5	32.8%	estimated ± 6.2 pp, medium confidence
74	Ornith-1.0-397B	32.7%	estimated ± 9.4 pp, low confidence
75	MiMo-V2.5	32.6%	estimated ± 8.3 pp, low confidence
76	GPT-5.1-Codex	31.8%	estimated ± 6.2 pp, medium confidence
77	GPT-5.2-Codex	31.7%	estimated ± 6.2 pp, medium confidence
78	Qwen3.7 Max	31.6%	measured
79	Inkling-Small	31.2%	measured
80	MiMo-V2.5-Pro	31.2%	measured
81	GLM-5.1	31.0%	measured
82	Solar Pro 4	30.5%	measured
83	Qwen 3.6 Max (preview)	30.1%	estimated ± 12.5 pp, low confidence
84	MiMo-V2-Pro	29.9%	estimated ± 8.3 pp, low confidence
85	Grok 4.3	29.2%	measured
86	Claude 4.1 Opus	29.0%	estimated ± 6.2 pp, medium confidence
87	Inkling	28.9%	measured
88	Hy3	28.3%	measured
89	Ornith-1.0-35B	27.9%	estimated ± 9.4 pp, low confidence
90	Solar Open 2	27.0%	estimated ± 8.0 pp, low confidence
91	Kimi K2.6	27.0%	measured
92	Kimi K2.7 Code	27.0%	measured
93	Grok 4.1	26.9%	estimated ± 8.3 pp, low confidence
94	Grok 4.1 Fast (Reasoning)	26.8%	estimated ± 12.5 pp, low confidence
95	Qwen3.5 Plus	26.7%	estimated ± 6.2 pp, medium confidence
96	Claude 4 Sonnet	26.6%	estimated ± 6.2 pp, medium confidence
97	Agents-A1-4B	26.3%	estimated ± 8.4 pp, medium confidence
98	GLM-4.7	25.8%	measured
99	GPT-5.4 mini	25.8%	measured
100	Step 3.7 Flash	25.8%	measured
101	Laguna S 2.1	25.7%	estimated ± 4.6 pp, medium confidence
102	MiniMax M2.7	25.7%	measured
103	Hy4 preview	25.4%	estimated ± 3.7 pp, low confidence
104	Muse Spark	25.1%	measured
105	Claude Haiku 4.5	25.0%	estimated ± 6.2 pp, medium confidence
106	Atria Dawn Preview	24.8%	estimated ± 3.7 pp, low confidence
107	Qwen3.8 Max	24.8%	estimated ± 3.7 pp, low confidence
108	Qwen3.6 Plus	24.7%	measured
109	Ornith-1.5-397B	24.6%	estimated ± 3.7 pp, low confidence
110	Qwen3.6-27B	24.4%	measured
111	Gemini 3.5 Flash-Lite	24.3%	measured
112	dots3-note Preview	24.0%	estimated ± 3.7 pp, low confidence
113	Ornith-1.0-9B	24.0%	estimated ± 9.4 pp, low confidence
114	Claude Opus 4.5	23.4%	estimated ± 3.7 pp, low confidence
115	A.X K2	23.2%	measured
116	Qwen3.5-27B	23.2%	estimated ± 8.4 pp, medium confidence
117	Qwen3.5-35B-A3B	23.2%	estimated ± 8.4 pp, medium confidence
118	Qwen3.5 397B	22.7%	estimated ± 3.7 pp, low confidence
119	GPT-5.4 nano	22.5%	measured
120	Ling 3.0 Flash	22.4%	measured
121	Ling 3.0 Flash FP8	22.4%	measured
122	Grok Build 0.1	22.3%	estimated ± 8.8 pp, medium confidence
123	Claude Opus 4.5 Thinking	22.3%	estimated ± 12.5 pp, low confidence
124	Gemini 3 Flash	21.9%	estimated ± 6.2 pp, medium confidence
125	Gemini 3 Pro	21.9%	estimated ± 6.2 pp, medium confidence
126	GLM-5	21.6%	estimated ± 3.7 pp, low confidence
127	BTL-4	21.5%	estimated ± 5.8 pp, medium confidence
128	Grok 4.1 Fast	21.0%	estimated ± 8.8 pp, medium confidence
129	Ornith-1.5-35B-A3B	21.0%	estimated ± 3.7 pp, low confidence
130	GPT-5 (high)	21.0%	measured
131	UI-Mate-9B	20.9%	estimated ± 8.8 pp, medium confidence
132	GLM-5-Turbo	20.2%	estimated ± 9.4 pp, low confidence
133	Qwen3.6-35B-A3B	19.8%	measured
134	LongCat-Flash-Lite-Sparse	19.8%	estimated ± 8.4 pp, medium confidence
135	GPT-5 (medium)	19.0%	estimated ± 12.5 pp, low confidence
136	Qwen3 Max	18.7%	estimated ± 8.8 pp, medium confidence
137	Ornith-1.5-9B	18.7%	estimated ± 3.7 pp, low confidence
138	Grok 4	17.9%	estimated ± 8.8 pp, medium confidence
139	Kimi K2.5	17.2%	measured
140	Kimi K2.5 (Reasoning)	17.2%	measured
141	GPT-5.1	16.5%	measured
142	Qwen3.5 397B (Reasoning)	16.5%	estimated ± 12.5 pp, low confidence
143	Qwen3.5-122B-A10B	15.8%	measured
144	GPT-5.1-Codex-Max	15.6%	estimated ± 12.5 pp, low confidence
145	Gemini 3.1 Flash-Lite	15.6%	estimated ± 8.8 pp, medium confidence
146	Grok 4.20	15.6%	estimated ± 8.8 pp, medium confidence
147	MiMo-V2-Omni	15.2%	estimated ± 9.4 pp, low confidence
148	Gemini 3.1 Pro	14.7%	measured
149	Muse Glimmer 30B	14.5%	measured
150	o3	13.6%	estimated ± 12.5 pp, low confidence
151	Qwen3.7 Plus	13.5%	measured
152	Pokee-Isaac 28B	13.3%	estimated ± 5.8 pp, medium confidence
153	Ternary Bonsai 2 27B	13.2%	estimated ± 12.5 pp, low confidence
154	Mistral Medium 3.5 128B	13.2%	measured
155	GLM-5V-Turbo	11.6%	estimated ± 8.8 pp, medium confidence
156	DeepSeek V3.2	11.0%	estimated ± 8.8 pp, low confidence
157	GLM-4.6	10.7%	estimated ± 12.5 pp, low confidence
158	MiniCPM5-2B	10.5%	measured
159	Grok Code Fast 1	9.9%	estimated ± 12.5 pp, low confidence
160	GPT-4.1	9.3%	estimated ± 8.8 pp, low confidence
161	Claude 4.1 Opus Thinking	7.3%	estimated ± 12.5 pp, low confidence
162	Nemotron 3.5 Lightning 30B A3B NVFP4	7.0%	measured
163	MiMo-V2-Flash	6.6%	measured
164	Gemma 4 31B	6.1%	measured
165	GPT-OSS 120B	5.6%	measured
166	Grok 4 Fast (Reasoning)	4.8%	estimated ± 12.5 pp, low confidence
167	Granite 4.2 30B	3.9%	measured
168	o1	3.6%	estimated ± 12.5 pp, low confidence
169	Gemma 4 26B A4B	3.4%	measured
170	Ling 3.0 Tiny	3.2%	measured
171	Kimi K2	3.2%	estimated ± 12.5 pp, low confidence
172	LLaDA2.2-flash	3.0%	estimated ± 5.8 pp, medium confidence
173	Command A+	0.9%	measured
174	Sarvam 105B	0.7%	estimated ± 12.5 pp, low confidence
175	GLM-4.5-Air	0.7%	estimated ± 12.5 pp, low confidence
176	DeepSeek V3.1 (Reasoning)	0.2%	estimated ± 12.5 pp, low confidence
177	DeepSeek-R1	0.2%	estimated ± 12.5 pp, low confidence
178	DeepSeek V3.1	0.1%	estimated ± 12.5 pp, low confidence
179	Sarvam 30B	0.1%	estimated ± 12.5 pp, low confidence
180	Solar Pro 2	0.1%	estimated ± 12.5 pp, low confidence
181	Mistral Large 2	0.1%	estimated ± 12.5 pp, low confidence
182	o3-mini	0.0%	estimated ± 12.5 pp, low confidence
183	Mistral Medium 3	0.0%	estimated ± 12.5 pp, low confidence
184	Qwen3-Omni-30B-A3B-Thinking	0.0%	estimated ± 12.5 pp, low confidence
185	Claude 3 Haiku	0.0%	estimated ± 12.5 pp, low confidence
186	Exaone 4.0 1.2B	0.0%	estimated ± 12.5 pp, low confidence
187	Granite-4.0-H-1B	0.0%	estimated ± 12.5 pp, low confidence
188	Llama 3.1 405B	0.0%	estimated ± 12.5 pp, low confidence
189	Qwen3-Omni-30B-A3B-Instruct	0.0%	estimated ± 12.5 pp, low confidence
190	Gemini 2.5 Flash	0.0%	estimated ± 12.5 pp, low confidence
191	Granite-4.0-H-350M	0.0%	estimated ± 12.5 pp, low confidence
192	Nova Pro	0.0%	estimated ± 12.5 pp, low confidence
193	Granite-4.0-350M	0.0%	estimated ± 12.5 pp, low confidence
194	Nemotron Ultra 253B	0.0%	estimated ± 12.5 pp, low confidence
195	LFM2.5-VL-1.6B-Extract	0.0%	estimated ± 12.5 pp, low confidence
196	Celeris-1	0.0%	measured
197	DeepSeek V3	0.0%	measured
198	DeepSeek V3 0324	0.0%	measured
199	Exaone 4.0 32B	0.0%	estimated ± 12.5 pp, low confidence
200	Gemini 2.5 Pro	0.0%	measured
201	Gemma 3 27B	0.0%	measured
202	Gemma 4 12B	0.0%	measured
203	Gemma 4 E2B	0.0%	measured
204	Gemma 4 E4B	0.0%	measured
205	GPT-4.1 mini	0.0%	measured
206	GPT-4.1 nano	0.0%	measured
207	GPT-4o	0.0%	measured
208	GPT-4o mini	0.0%	measured
209	GPT-OSS 20B	0.0%	measured
210	Granite 4.2 3B	0.0%	measured
211	Granite 4.2 8B	0.0%	measured
212	K-Exaone	0.0%	measured
213	LFM2.5-2.6B	0.0%	measured
214	LFM2.5-230M	0.0%	estimated ± 5.8 pp, low confidence
215	LFM2.5-8B-A1B	0.0%	estimated ± 5.8 pp, low confidence
216	LFM2.5-VL-3B	0.0%	estimated ± 5.8 pp, low confidence
217	LFM2.5-VL-450M	0.0%	estimated ± 5.8 pp, low confidence
218	Ling 2.6 Flash	0.0%	measured
219	LLaDA2.2-mini	0.0%	estimated ± 5.8 pp, low confidence
220	Llama 4 Maverick	0.0%	measured
221	Llama 4 Scout	0.0%	measured
222	Mellum2-12B-A2.5B-Instruct	0.0%	estimated ± 5.8 pp, low confidence
223	Mellum2-12B-A2.5B-Thinking	0.0%	estimated ± 5.8 pp, low confidence
224	Mercury 2.5	0.0%	measured
225	MiniCPM5-1B	0.0%	estimated ± 5.8 pp, low confidence
226	Mistral Large 3	0.0%	measured
227	Mistral Small 4	0.0%	measured
228	Mistral Small 4 (Reasoning)	0.0%	measured
229	Nemotron 3 Nano 30B	0.0%	measured
230	Nemotron 3 Nano Omni 30B A3B	0.0%	measured
231	Nemotron 3 Super 100B	0.0%	measured
232	North Mini Code	0.0%	measured
233	Phi-4	0.0%	estimated ± 12.5 pp, low confidence
234	Solar Pro 3	0.0%	measured
235	Trinity-Large-Preview	0.0%	measured
236	Trinity-Large-Thinking	0.0%	measured
237	Ultravox v0.6 Llama 3.3 70B	0.0%	measured
238	ZAYA1-8B	0.0%	estimated ± 5.8 pp, low confidence

GDPval-AA leaderboard

All leaderboards

Agentic · terminal

Agentic · tools

Coding

Math

Knowledge & reasoning

Instruction following

Multilingual

Vision & documents

Long context

General