Agentic · tools

APEX-Agents leaderboard

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

As of 2026-10-07, the highest measured score on APEX-Agents is 57.5% by Grok 4.6. 125 more models have estimated scores, calibrated from the benchmarks they were measured on.

Measured scores: benchlm.ai.

#	Model	Score	Source
1	Grok 4.6	57.5%	measured
2	Claude Opus 5.5	52.3%	estimated ± 10.0 pp, low confidence
3	Claude Sonnet 5.5	51.5%	estimated ± 10.0 pp, low confidence
4	Claude Fable 5.1	49.1%	estimated ± 10.0 pp, low confidence
5	Grok 4.7	47.9%	estimated ± 10.0 pp, low confidence
6	MiMo-V2.6-Pro	47.0%	estimated ± 10.0 pp, low confidence
7	Muse Spark 1.3	46.9%	estimated ± 10.0 pp, low confidence
8	Qwen3.8 Max Preview	46.6%	estimated ± 10.0 pp, low confidence
9	GLM-5.3	46.0%	estimated ± 10.0 pp, low confidence
10	Qwen3.8-Flash-Next	45.3%	estimated ± 10.0 pp, low confidence
11	Gemini 4 Argon	45.1%	estimated ± 10.0 pp, low confidence
12	Ling 3.1 Flash	45.0%	estimated ± 10.0 pp, low confidence
13	Claude Fable 5	44.7%	estimated ± 10.0 pp, low confidence
14	GPT-5.6 Sol	44.7%	estimated ± 10.0 pp, low confidence
15	MiMo-V2.6-Flash	44.6%	estimated ± 10.0 pp, low confidence
16	DeepSeek V4.1 Flash	44.3%	estimated ± 10.0 pp, low confidence
17	GPT-6.1 Sol	43.6%	estimated ± 10.0 pp, low confidence
18	GPT-6 Astra	42.5%	estimated ± 10.0 pp, low confidence
19	GPT-6 Sol	41.5%	estimated ± 10.0 pp, low confidence
20	Muse Spark 1.2	40.4%	estimated ± 10.0 pp, low confidence
21	Muse Spark 1.1	40.4%	estimated ± 0.6 pp, low confidence
22	Claude Sonnet 5	40.0%	estimated ± 10.0 pp, low confidence
23	GPT-5.6 Luna	39.9%	estimated ± 10.0 pp, low confidence
24	GPT-5.6 Terra	39.6%	estimated ± 10.0 pp, low confidence
25	GPT-6 Luna	39.1%	estimated ± 10.0 pp, low confidence
26	Gemini 3.8 Flash	39.0%	estimated ± 10.0 pp, low confidence
27	Mistral Large 4	38.6%	estimated ± 10.0 pp, low confidence
28	Qwen3.8-27B	38.6%	estimated ± 10.0 pp, low confidence
29	Claude Opus 5	38.5%	estimated ± 0.6 pp, low confidence
30	Apodex 1.1	38.5%	measured
31	Step 5 Preview	37.8%	measured
32	Kimi K3	37.6%	measured
33	Gemini 3.7 Flash	37.5%	estimated ± 10.0 pp, low confidence
34	Grok 4.5	37.5%	estimated ± 10.0 pp, low confidence
35	Hy4 preview	37.1%	measured
36	Gemini 3.5 Flash	36.8%	estimated ± 0.6 pp, medium confidence
37	Claude Opus 4.8	35.6%	estimated ± 0.6 pp, medium confidence
38	Ornith-1.5-397B	33.9%	estimated ± 0.6 pp, medium confidence
39	Gemini 3.6 Flash	33.8%	estimated ± 10.0 pp, low confidence
40	Inkling-Small	33.6%	estimated ± 0.6 pp, medium confidence
41	Beam	32.8%	estimated ± 0.6 pp, medium confidence
42	Claude Opus 4.7 (Adaptive)	31.7%	estimated ± 0.6 pp, medium confidence
43	GLM-5.2	31.3%	estimated ± 0.6 pp, medium confidence
44	Hy3 Preview	31.3%	estimated ± 10.0 pp, low confidence
45	Qwen3.7 Max	31.0%	estimated ± 0.6 pp, medium confidence
46	dots3-note Preview	30.8%	measured
47	Kimi K2.7 Code	30.7%	estimated ± 0.6 pp, medium confidence
48	Muse Glimmer 30B	30.3%	estimated ± 0.6 pp, medium confidence
49	GPT-5.5	30.1%	estimated ± 0.6 pp, medium confidence
50	Quasar 438B	30.1%	estimated ± 10.0 pp, low confidence
51	Ling 3.0 Flash VL	29.3%	estimated ± 10.0 pp, low confidence
52	Nemotron 3 Ultra	29.3%	estimated ± 10.0 pp, low confidence
53	MiniMax M3	29.2%	estimated ± 0.6 pp, medium confidence
54	Ling 3.0 Flash Fin	29.2%	measured
55	Inkling	29.1%	estimated ± 0.6 pp, medium confidence
56	DeepSeek V4 Pro 0813	28.7%	estimated ± 0.6 pp, medium confidence
57	Qwen3.7 Plus	28.4%	estimated ± 0.6 pp, medium confidence
58	MiMo-V2.5-Pro	27.8%	estimated ± 10.0 pp, low confidence
59	Apodex 1.1 Mini	27.7%	measured
60	GLM-5.1	27.3%	estimated ± 0.6 pp, medium confidence
61	GPT-5.4	26.3%	estimated ± 0.6 pp, medium confidence
62	Grok 4.3	26.3%	estimated ± 10.0 pp, low confidence
63	Ornith-1.5-35B-A3B	26.0%	estimated ± 0.6 pp, medium confidence
64	Hy3	25.6%	estimated ± 10.0 pp, low confidence
65	DeepSeek V4 Flash 0731	25.1%	estimated ± 0.6 pp, medium confidence
66	GLM-4.7	23.6%	estimated ± 10.0 pp, low confidence
67	Step 3.7 Flash	23.6%	estimated ± 10.0 pp, low confidence
68	MiniMax M2.7	23.5%	estimated ± 10.0 pp, low confidence
69	Muse Spark	23.0%	estimated ± 10.0 pp, low confidence
70	Qwen3.6-27B	22.4%	estimated ± 10.0 pp, low confidence
71	Gemini 3.5 Flash-Lite	22.4%	estimated ± 10.0 pp, low confidence
72	Ling 3.0 Flash	22.3%	estimated ± 0.6 pp, medium confidence
73	A.X K2	21.5%	estimated ± 10.0 pp, low confidence
74	Ling 3.0 Flash FP8	20.8%	estimated ± 10.0 pp, low confidence
75	Qwen3.6-35B-A3B	20.1%	estimated ± 0.6 pp, medium confidence
76	GPT-5 (high)	19.6%	estimated ± 10.0 pp, low confidence
77	Solar Pro 4	18.7%	measured
78	Solar Open 2	16.6%	measured
79	Kimi K2.5 (Reasoning)	16.4%	estimated ± 10.0 pp, low confidence
80	GPT-5.4 mini	16.0%	estimated ± 0.6 pp, low confidence
81	GPT-5.1	15.7%	estimated ± 10.0 pp, low confidence
82	Qwen3.5-122B-A10B	15.1%	estimated ± 10.0 pp, low confidence
83	GPT-5.4 nano	14.7%	estimated ± 0.6 pp, low confidence
84	Kimi K2.6	14.6%	estimated ± 0.6 pp, low confidence
85	Gemini 3.1 Pro	14.1%	estimated ± 10.0 pp, low confidence
86	Ornith-1.5-9B	13.2%	estimated ± 0.6 pp, low confidence
87	Mistral Medium 3.5 128B	12.8%	estimated ± 10.0 pp, low confidence
88	MiniCPM5-2B	10.3%	estimated ± 10.0 pp, low confidence
89	Qwen3.6 Plus	8.4%	estimated ± 0.6 pp, low confidence
90	Nemotron 3.5 Lightning 30B A3B NVFP4	7.0%	estimated ± 10.0 pp, low confidence
91	LLaDA2.2-flash	6.8%	estimated ± 0.6 pp, low confidence
92	Qwen3.5 397B	6.7%	estimated ± 0.6 pp, low confidence
93	MiMo-V2-Flash	6.6%	estimated ± 10.0 pp, low confidence
94	LongCat-Flash-Lite-Sparse	6.3%	estimated ± 0.6 pp, low confidence
95	Gemma 4 31B	6.1%	estimated ± 10.0 pp, low confidence
96	GPT-OSS 120B	5.6%	estimated ± 10.0 pp, low confidence
97	Granite 4.2 30B	4.0%	estimated ± 10.0 pp, low confidence
98	Claude Opus 4.5	3.7%	estimated ± 0.6 pp, low confidence
99	Gemma 4 26B A4B	3.5%	estimated ± 10.0 pp, low confidence
100	Ling 3.0 Tiny	3.3%	estimated ± 10.0 pp, low confidence
101	Command A+	0.9%	estimated ± 10.0 pp, low confidence
102	Celeris-1	0.0%	estimated ± 10.0 pp, low confidence
103	DeepSeek V3	0.0%	estimated ± 10.0 pp, low confidence
104	DeepSeek V3 0324	0.0%	estimated ± 10.0 pp, low confidence
105	Gemini 2.5 Pro	0.0%	estimated ± 10.0 pp, low confidence
106	Gemma 3 27B	0.0%	estimated ± 10.0 pp, low confidence
107	Gemma 4 12B	0.0%	estimated ± 10.0 pp, low confidence
108	Gemma 4 E2B	0.0%	estimated ± 10.0 pp, low confidence
109	Gemma 4 E4B	0.0%	estimated ± 10.0 pp, low confidence
110	GLM-5	0.0%	estimated ± 0.6 pp, low confidence
111	GPT-4.1 mini	0.0%	estimated ± 10.0 pp, low confidence
112	GPT-4.1 nano	0.0%	estimated ± 10.0 pp, low confidence
113	GPT-4o	0.0%	estimated ± 10.0 pp, low confidence
114	GPT-4o mini	0.0%	estimated ± 10.0 pp, low confidence
115	GPT-OSS 20B	0.0%	estimated ± 10.0 pp, low confidence
116	Granite 4.2 3B	0.0%	estimated ± 10.0 pp, low confidence
117	Granite 4.2 8B	0.0%	estimated ± 10.0 pp, low confidence
118	K-Exaone	0.0%	estimated ± 10.0 pp, low confidence
119	Kimi K2.5	0.0%	estimated ± 0.6 pp, low confidence
120	LFM2.5-2.6B	0.0%	estimated ± 10.0 pp, low confidence
121	Ling 2.6 Flash	0.0%	estimated ± 10.0 pp, low confidence
122	Llama 4 Maverick	0.0%	estimated ± 10.0 pp, low confidence
123	Llama 4 Scout	0.0%	estimated ± 10.0 pp, low confidence
124	Mercury 2.5	0.0%	estimated ± 10.0 pp, low confidence
125	Mistral Large 3	0.0%	estimated ± 10.0 pp, low confidence
126	Mistral Small 4	0.0%	estimated ± 10.0 pp, low confidence
127	Mistral Small 4 (Reasoning)	0.0%	estimated ± 10.0 pp, low confidence
128	Nemotron 3 Nano 30B	0.0%	estimated ± 10.0 pp, low confidence
129	Nemotron 3 Nano Omni 30B A3B	0.0%	estimated ± 10.0 pp, low confidence
130	Nemotron 3 Super 100B	0.0%	estimated ± 10.0 pp, low confidence
131	North Mini Code	0.0%	estimated ± 10.0 pp, low confidence
132	Solar Pro 3	0.0%	estimated ± 10.0 pp, low confidence
133	Trinity-Large-Preview	0.0%	estimated ± 10.0 pp, low confidence
134	Trinity-Large-Thinking	0.0%	estimated ± 10.0 pp, low confidence
135	Ultravox v0.6 Llama 3.3 70B	0.0%	estimated ± 10.0 pp, low confidence

APEX-Agents leaderboard

All leaderboards

Agentic · terminal

Agentic · tools

Coding

Math

Knowledge & reasoning

Instruction following

Multilingual

Vision & documents

Long context

General