Agentic · tools

Gert Labs leaderboard

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

As of 2026-10-07, the highest measured score on Gert Labs is 73.0% by Claude Opus 4.8. 133 more models have estimated scores, calibrated from the benchmarks they were measured on.

Measured scores: benchlm.ai.

#	Model	Score	Source
1	Claude Fable 5.1	100.0%	estimated ± 3.0 pp, low confidence
2	GPT-6 Astra	100.0%	estimated ± 3.0 pp, low confidence
3	Claude Sonnet 5.5	85.6%	estimated ± 8.9 pp, low confidence
4	Qwen3.8 Max Preview	81.6%	estimated ± 8.5 pp, low confidence
5	Claude Opus 5	81.4%	estimated ± 3.0 pp, low confidence
6	Grok 4.7	81.3%	estimated ± 8.9 pp, low confidence
7	Claude Fable 5	80.2%	estimated ± 3.0 pp, low confidence
8	Qwen3.8-27B	78.5%	estimated ± 7.2 pp, low confidence
9	Gemini 4 Argon	76.7%	estimated ± 3.7 pp, low confidence
10	Gemini 3.6 Flash	76.4%	estimated ± 7.2 pp, medium confidence
11	GPT-6 Sol	76.4%	estimated ± 3.7 pp, low confidence
12	Muse Spark 1.2	75.9%	estimated ± 8.5 pp, low confidence
13	GPT-6.1 Sol	75.9%	estimated ± 8.9 pp, low confidence
14	GPT-5.5 Pro	75.9%	estimated ± 7.4 pp, low confidence
15	Claude Opus 5.5	75.8%	estimated ± 3.7 pp, low confidence
16	Holo3-35B-A3B	75.7%	estimated ± 7.2 pp, medium confidence
17	GPT-5.6 Sol	75.5%	estimated ± 3.0 pp, low confidence
18	GPT-5.4 Pro	74.7%	estimated ± 7.4 pp, low confidence
19	Gemini 3.8 Flash	74.4%	estimated ± 3.0 pp, low confidence
20	Grok 4.5	74.0%	estimated ± 8.5 pp, medium confidence
21	MiMo-V2.6-Flash	73.4%	estimated ± 6.0 pp, low confidence
22	MiMo-V2.6-Pro	73.0%	estimated ± 6.0 pp, low confidence
23	Claude Opus 4.8	73.0%	measured
24	GPT-5.5	72.9%	measured
25	Qwen3.8-Flash-Next	71.9%	estimated ± 3.7 pp, medium confidence
26	Qwen3.8 Max	71.9%	estimated ± 3.7 pp, medium confidence
27	Muse Spark 1.3	71.5%	estimated ± 3.0 pp, medium confidence
28	Claude Opus 4.7 (Adaptive)	71.5%	estimated ± 3.7 pp, medium confidence
29	DeepSeek V4.1 Flash	70.9%	estimated ± 6.0 pp, low confidence
30	Kimi K3	70.9%	estimated ± 3.0 pp, medium confidence
31	Ling 3.1 Flash	70.8%	estimated ± 6.0 pp, low confidence
32	Fugu Cyber	70.5%	estimated ± 6.0 pp, low confidence
33	Atria Dawn Preview	70.3%	estimated ± 6.0 pp, low confidence
34	Gemini 3.8 Flash Cyber	70.2%	estimated ± 6.0 pp, low confidence
35	Holo3-122B-A10B	70.1%	estimated ± 7.2 pp, medium confidence
36	GPT-6 Luna	70.0%	estimated ± 8.9 pp, medium confidence
37	Claude Sonnet 5	69.7%	estimated ± 3.0 pp, medium confidence
38	Gemini 3.7 Flash	69.7%	estimated ± 3.0 pp, medium confidence
39	GPT-5.6 Terra	69.7%	estimated ± 3.0 pp, medium confidence
40	Step 5 Preview	69.7%	estimated ± 6.0 pp, low confidence
41	Muse Spark 1.1	69.7%	estimated ± 3.7 pp, medium confidence
42	GLM-5.3	69.6%	estimated ± 6.0 pp, low confidence
43	Claude Mythos 5	69.3%	estimated ± 6.0 pp, low confidence
44	Mistral Large 4	69.3%	estimated ± 8.9 pp, medium confidence
45	DeepSeek V4 Pro 0813	69.2%	estimated ± 6.0 pp, low confidence
46	Gemini 3.5 Flash Cyber	69.1%	estimated ± 6.0 pp, low confidence
47	Claude Mythos Preview	69.1%	estimated ± 6.0 pp, low confidence
48	Grok 4.6	68.0%	estimated ± 3.0 pp, medium confidence
49	UI-Mate-27B	67.4%	estimated ± 7.2 pp, medium confidence
50	Hy4 preview	67.4%	estimated ± 6.0 pp, low confidence
51	DeepSeek V4 Flash 0731	66.8%	estimated ± 6.0 pp, low confidence
52	Claude Opus 4.7	65.6%	measured
53	GPT-5.4	64.9%	measured
54	Quasar 438B	64.6%	estimated ± 8.5 pp, medium confidence
55	GPT-5.6 Luna	64.5%	estimated ± 3.0 pp, medium confidence
56	Qwen3.7 Max	64.3%	measured
57	Claude Opus 4.5	64.2%	measured
58	Kimi K2.7 Code	63.9%	estimated ± 8.2 pp, low confidence
59	Gemini 3.5 Flash-Lite	63.3%	estimated ± 7.2 pp, medium confidence
60	Gemini 3 Pro	63.2%	measured
61	Claude Sonnet 4.6	62.9%	measured
62	K-EXAONE 2.0	62.7%	estimated ± 9.1 pp, low confidence
63	MiMo-V2.5-Pro	62.7%	measured
64	Ornith-1.0-397B	62.4%	estimated ± 9.1 pp, low confidence
65	Claude Opus 4.6	61.9%	measured
66	Gemini 3.5 Flash	61.9%	measured
67	GLM-5.1	60.1%	measured
68	Claude Opus 4.6 (Adaptive)	59.5%	estimated ± 11.3 pp, low confidence
69	Inkling-Small	58.5%	estimated ± 7.4 pp, medium confidence
70	Beam	58.5%	estimated ± 7.4 pp, medium confidence
71	Apodex 1.1	58.3%	estimated ± 8.9 pp, medium confidence
72	Apodex 1.1 Mini	58.3%	estimated ± 8.9 pp, medium confidence
73	Ornith-1.0-35B	58.2%	estimated ± 9.1 pp, low confidence
74	Inkling	58.1%	estimated ± 7.4 pp, medium confidence
75	Solar Open 2	57.5%	estimated ± 8.2 pp, low confidence
76	Hy3	57.5%	estimated ± 8.5 pp, medium confidence
77	GPT-5.3 Codex	57.5%	measured
78	Gemini 3.1 Pro	56.9%	measured
79	Kimi K2.6	56.8%	measured
80	Ling 3.0 Flash VL	56.7%	estimated ± 8.9 pp, medium confidence
81	Gemini 3 Flash	56.6%	measured
82	MiniMax M3	56.2%	estimated ± 3.7 pp, medium confidence
83	Agents-A1	56.2%	estimated ± 7.4 pp, medium confidence
84	Ornith-1.5-397B	55.5%	estimated ± 6.1 pp, low confidence
85	GLM-5.2	55.3%	estimated ± 7.2 pp, medium confidence
86	Qwen3.6-27B	54.8%	measured
87	Muse Spark	54.8%	estimated ± 6.0 pp, low confidence
88	dots3-note Preview	54.5%	estimated ± 6.1 pp, low confidence
89	Ornith-1.0-9B	54.1%	estimated ± 9.1 pp, low confidence
90	UI-Mate-9B	53.3%	estimated ± 7.2 pp, medium confidence
91	LLaDA2.2-flash	53.2%	estimated ± 8.2 pp, low confidence
92	Muse Glimmer 30B	52.9%	estimated ± 7.2 pp, medium confidence
93	Ling 3.0 Flash FP8	52.8%	estimated ± 8.5 pp, medium confidence
94	Ling 3.0 Flash	51.8%	estimated ± 6.1 pp, low confidence
95	GPT-5.2-Codex	51.8%	measured
96	Step 3.7 Flash	51.6%	measured
97	GLM-5	51.0%	measured
98	Claude 4.1 Opus	50.8%	estimated ± 7.3 pp, medium confidence
99	GPT-5.4 mini	50.7%	estimated ± 7.2 pp, medium confidence
100	Qwen3.6 Plus	50.6%	measured
101	Qwen3.5 Plus	50.4%	estimated ± 7.3 pp, medium confidence
102	LLaDA2.2-mini	50.3%	estimated ± 9.1 pp, low confidence
103	Claude Haiku 4.5	50.3%	estimated ± 7.3 pp, medium confidence
104	GPT-5 (high)	50.2%	estimated ± 7.3 pp, low confidence
105	GPT-5.1-Codex	49.7%	measured
106	GLM-5-Turbo	49.4%	estimated ± 9.1 pp, low confidence
107	Grok Build 0.1	49.2%	measured
108	Ornith-1.5-35B-A3B	48.7%	estimated ± 6.1 pp, low confidence
109	Claude Sonnet 4.5	48.5%	measured
110	Grok 4.1	48.3%	estimated ± 8.5 pp, low confidence
111	Qwen3.7 Plus	47.5%	estimated ± 3.7 pp, low confidence
112	Grok 4.1 Fast	47.3%	measured
113	GPT-5.4 nano	47.2%	estimated ± 7.2 pp, medium confidence
114	A.X K2	47.2%	estimated ± 8.9 pp, medium confidence
115	MiMo-V2.5	46.9%	measured
116	Qwen3.5 397B	46.8%	measured
117	GPT-5.2	46.5%	measured
118	Agents-A1-4B	46.4%	estimated ± 7.4 pp, medium confidence
119	Kimi K2.5	45.9%	measured
120	Ornith-1.5-9B	44.1%	estimated ± 6.1 pp, low confidence
121	Qwen3.5-122B-A10B	44.0%	estimated ± 7.2 pp, medium confidence
122	Grok 4.3	43.9%	measured
123	Qwen3 Max	43.7%	measured
124	Qwen3.6-35B-A3B	42.7%	measured
125	Grok 4	42.3%	measured
126	Gemini 2.5 Pro	42.0%	measured
127	MiMo-V2-Omni	42.0%	estimated ± 9.1 pp, low confidence
128	GPT-5.1	41.2%	measured
129	MiniMax M2.7	40.4%	measured
130	GLM-4.7	40.0%	measured
131	Claude 4 Sonnet	39.7%	measured
132	Qwen3.5-27B	39.4%	measured
133	Mistral Medium 3.5 128B	39.1%	measured
134	Gemini 3.1 Flash-Lite	38.5%	measured
135	Grok 4.20	38.4%	measured
136	MiniCPM5-2B	38.3%	estimated ± 8.9 pp, medium confidence
137	Hy3 Preview	36.9%	measured
138	MiMo-V2-Flash	36.9%	estimated ± 8.9 pp, medium confidence
139	MiMo-V2-Pro	36.7%	measured
140	Granite 4.2 30B	36.3%	estimated ± 8.9 pp, medium confidence
141	Gemma 4 26B A4B	36.2%	estimated ± 8.9 pp, medium confidence
142	Ling 3.0 Tiny	36.2%	estimated ± 8.9 pp, medium confidence
143	DeepSeek V3 0324	36.0%	estimated ± 8.9 pp, medium confidence
144	Gemma 4 12B	36.0%	estimated ± 8.9 pp, medium confidence
145	Gemma 4 E2B	36.0%	estimated ± 8.9 pp, medium confidence
146	Gemma 4 E4B	36.0%	estimated ± 8.9 pp, medium confidence
147	GPT-4.1 mini	36.0%	estimated ± 8.9 pp, medium confidence
148	GPT-4.1 nano	36.0%	estimated ± 8.9 pp, medium confidence
149	GPT-4o	36.0%	estimated ± 8.9 pp, medium confidence
150	GPT-4o mini	36.0%	estimated ± 8.9 pp, medium confidence
151	Granite 4.2 3B	36.0%	estimated ± 8.9 pp, medium confidence
152	K-Exaone	36.0%	estimated ± 8.9 pp, medium confidence
153	LFM2.5-2.6B	36.0%	estimated ± 8.9 pp, medium confidence
154	Ling 2.6 Flash	36.0%	estimated ± 8.9 pp, medium confidence
155	Mercury 2.5	36.0%	estimated ± 8.9 pp, medium confidence
156	Nemotron 3 Nano Omni 30B A3B	36.0%	estimated ± 8.9 pp, medium confidence
157	North Mini Code	36.0%	estimated ± 8.9 pp, medium confidence
158	Solar Pro 3	36.0%	estimated ± 8.9 pp, medium confidence
159	Ultravox v0.6 Llama 3.3 70B	36.0%	estimated ± 8.9 pp, medium confidence
160	Nemotron 3 Super 100B	35.8%	estimated ± 8.5 pp, medium confidence
161	Granite 4.2 8B	35.4%	estimated ± 8.5 pp, medium confidence
162	Command A+	35.3%	estimated ± 8.5 pp, medium confidence
163	Gemma 4 31B	35.3%	measured
164	Mistral Large 3	34.1%	estimated ± 8.5 pp, medium confidence
165	Mistral Small 4	33.1%	estimated ± 8.5 pp, medium confidence
166	Mistral Small 4 (Reasoning)	33.1%	estimated ± 8.5 pp, medium confidence
167	GPT-OSS 20B	33.1%	estimated ± 8.5 pp, medium confidence
168	Trinity-Large-Preview	32.9%	estimated ± 8.5 pp, medium confidence
169	Nemotron 3 Nano 30B	32.7%	estimated ± 8.5 pp, low confidence
170	Kimi K2.5 (Reasoning)	32.6%	measured
171	Trinity-Large-Thinking	32.6%	measured
172	DeepSeek V3	32.5%	estimated ± 8.5 pp, low confidence
173	Celeris-1	32.3%	estimated ± 8.5 pp, low confidence
174	Llama 4 Maverick	32.3%	estimated ± 8.5 pp, low confidence
175	Llama 4 Scout	32.2%	estimated ± 8.5 pp, low confidence
176	Gemma 3 27B	31.8%	estimated ± 8.5 pp, low confidence
177	GLM-5V-Turbo	30.8%	measured
178	Solar Pro 4	30.2%	estimated ± 7.4 pp, low confidence
179	LongCat-Flash-Lite-Sparse	29.7%	estimated ± 7.4 pp, low confidence
180	GPT-OSS 120B	29.6%	measured
181	DeepSeek V3.2	29.6%	measured
182	Qwen3.5-35B-A3B	29.0%	measured
183	Nemotron 3 Ultra	26.4%	estimated ± 7.4 pp, low confidence
184	GPT-4.1	25.7%	measured
185	Nemotron 3.5 Lightning 30B A3B NVFP4	20.9%	estimated ± 7.4 pp, low confidence

Gert Labs leaderboard

All leaderboards

Agentic · terminal

Agentic · tools

Coding

Math

Knowledge & reasoning

Instruction following

Multilingual

Vision & documents

Long context

General