Agentic · tools

ApprenticeBench leaderboard

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

As of 2026-10-07, the highest measured score on ApprenticeBench is 72.0% by Claude Fable 5.1. 94 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	72.0%	measured
2	GPT-6 Astra	68.0%	measured
3	Claude Opus 5	36.0%	measured
4	Claude Fable 5	34.0%	measured
5	Qwen3.8 Max	33.3%	estimated ± 5.2 pp, low confidence
6	MiMo-V2.6-Flash	32.7%	estimated ± 4.0 pp, low confidence
7	MiMo-V2.6-Pro	32.6%	estimated ± 4.0 pp, low confidence
8	DeepSeek V4.1 Flash	31.0%	estimated ± 4.0 pp, low confidence
9	Ling 3.1 Flash	30.9%	estimated ± 4.0 pp, low confidence
10	Fugu Cyber	30.0%	estimated ± 4.0 pp, low confidence
11	Atria Dawn Preview	29.5%	estimated ± 4.0 pp, low confidence
12	Qwen3.8-27B	29.5%	estimated ± 5.2 pp, low confidence
13	Gemini 3.8 Flash Cyber	29.1%	estimated ± 4.0 pp, low confidence
14	Gemini 3.6 Flash	26.7%	estimated ± 5.2 pp, low confidence
15	Step 5 Preview	26.4%	estimated ± 4.0 pp, low confidence
16	GPT-5.6 Sol	26.0%	measured
17	GLM-5.3	25.9%	estimated ± 4.0 pp, medium confidence
18	Holo3-35B-A3B	25.7%	estimated ± 5.2 pp, low confidence
19	Claude Mythos 5	24.2%	estimated ± 4.0 pp, medium confidence
20	Gemini 3.8 Flash	24.0%	measured
21	dots3-note Preview	23.1%	estimated ± 10.1 pp, low confidence
22	DeepSeek V4 Pro 0813	22.8%	estimated ± 4.0 pp, medium confidence
23	Gemini 3.5 Flash Cyber	22.5%	estimated ± 4.0 pp, medium confidence
24	Claude Mythos Preview	22.2%	estimated ± 4.0 pp, medium confidence
25	Qwen3.8 Max Preview	20.3%	estimated ± 5.0 pp, low confidence
26	GPT-5.5	20.0%	measured
27	Claude Opus 4.8	20.0%	estimated ± 3.1 pp, low confidence
28	Qwen3.8-Flash-Next	19.0%	estimated ± 5.6 pp, low confidence
29	Muse Spark 1.3	19.0%	measured
30	Kimi K3	18.0%	measured
31	Holo3-122B-A10B	17.7%	estimated ± 5.2 pp, low confidence
32	Claude Sonnet 5	16.0%	measured
33	Gemini 3.7 Flash	16.0%	measured
34	GPT-5.6 Terra	16.0%	measured
35	UI-Mate-27B	13.7%	estimated ± 5.2 pp, low confidence
36	Grok 4.6	13.0%	measured
37	Beam	11.8%	estimated ± 10.1 pp, low confidence
38	GPT-5.4	11.0%	measured
39	GLM-5.2	10.0%	estimated ± 5.0 pp, low confidence
40	Claude Opus 4.6 (Adaptive)	9.2%	estimated ± 5.0 pp, low confidence
41	Hy4 preview	8.8%	estimated ± 4.0 pp, medium confidence
42	Gemini 3.5 Flash-Lite	7.2%	estimated ± 5.2 pp, low confidence
43	Apodex 1.1	7.0%	estimated ± 5.0 pp, low confidence
44	Apodex 1.1 Mini	7.0%	estimated ± 5.0 pp, low confidence
45	Claude Opus 4.7	7.0%	measured
46	GPT-5.6 Luna	7.0%	measured
47	Qwen3.7 Max	7.0%	estimated ± 3.1 pp, medium confidence
48	Claude Opus 4.5	6.9%	estimated ± 3.1 pp, medium confidence
49	DeepSeek V4 Flash 0731	6.2%	estimated ± 4.0 pp, medium confidence
50	Gemini 3 Pro	5.4%	estimated ± 3.1 pp, medium confidence
51	Claude Opus 4.6	5.0%	measured
52	MiMo-V2.5-Pro	4.6%	estimated ± 3.1 pp, medium confidence
53	Claude Opus 4.7 (Adaptive)	4.0%	estimated ± 4.0 pp, medium confidence
54	GPT-5.4 mini	3.5%	estimated ± 5.0 pp, low confidence
55	Muse Spark 1.1	3.4%	estimated ± 4.0 pp, low confidence
56	Muse Spark	3.4%	estimated ± 4.0 pp, low confidence
57	Gemini 3.5 Flash	3.3%	estimated ± 3.1 pp, medium confidence
58	Claude Sonnet 4.6	2.0%	measured
59	Mercury 2.5	1.3%	estimated ± 10.1 pp, low confidence
60	GLM-5.1	0.7%	estimated ± 3.1 pp, low confidence
61	Claude 4.1 Opus	0.5%	estimated ± 5.6 pp, low confidence
62	Qwen3.5 Plus	0.2%	estimated ± 5.6 pp, low confidence
63	Claude Haiku 4.5	0.1%	estimated ± 5.6 pp, low confidence
64	GPT-5 (high)	0.0%	estimated ± 5.6 pp, low confidence
65	Claude 4 Sonnet	0.0%	estimated ± 3.1 pp, low confidence
66	Claude Sonnet 4.5	0.0%	estimated ± 3.1 pp, low confidence
67	DeepSeek V3.2	0.0%	estimated ± 3.1 pp, low confidence
68	Gemini 2.5 Pro	0.0%	estimated ± 3.1 pp, low confidence
69	Gemini 3.1 Flash-Lite	0.0%	estimated ± 3.1 pp, low confidence
70	Gemini 3.1 Pro	0.0%	estimated ± 3.1 pp, low confidence
71	Gemini 3 Flash	0.0%	estimated ± 3.1 pp, low confidence
72	Gemma 4 31B	0.0%	estimated ± 3.1 pp, low confidence
73	GLM-4.7	0.0%	estimated ± 3.1 pp, low confidence
74	GLM-5	0.0%	estimated ± 3.1 pp, low confidence
75	GLM-5V-Turbo	0.0%	estimated ± 3.1 pp, low confidence
76	GPT-4.1	0.0%	estimated ± 3.1 pp, low confidence
77	GPT-5.1	0.0%	estimated ± 3.1 pp, low confidence
78	GPT-5.1-Codex	0.0%	estimated ± 3.1 pp, low confidence
79	GPT-5.2	0.0%	estimated ± 3.1 pp, low confidence
80	GPT-5.2-Codex	0.0%	estimated ± 3.1 pp, low confidence
81	GPT-5.3 Codex	0.0%	estimated ± 3.1 pp, low confidence
82	GPT-5.4 nano	0.0%	estimated ± 5.0 pp, low confidence
83	GPT-OSS 120B	0.0%	estimated ± 3.1 pp, low confidence
84	GPT-OSS 20B	0.0%	estimated ± 5.0 pp, low confidence
85	Grok 4	0.0%	estimated ± 3.1 pp, low confidence
86	Grok 4.1 Fast	0.0%	estimated ± 3.1 pp, low confidence
87	Grok 4.20	0.0%	estimated ± 3.1 pp, low confidence
88	Grok 4.3	0.0%	estimated ± 3.1 pp, low confidence
89	Grok Build 0.1	0.0%	estimated ± 3.1 pp, low confidence
90	Hy3 Preview	0.0%	estimated ± 3.1 pp, low confidence
91	Kimi K2.6	0.0%	estimated ± 3.1 pp, low confidence
92	Kimi K2.5	0.0%	estimated ± 3.1 pp, low confidence
93	Kimi K2.5 (Reasoning)	0.0%	estimated ± 3.1 pp, low confidence
94	MiMo-V2.5	0.0%	estimated ± 3.1 pp, low confidence
95	MiMo-V2-Pro	0.0%	estimated ± 3.1 pp, low confidence
96	MiniMax M2.7	0.0%	estimated ± 3.1 pp, low confidence
97	MiniMax M3	0.0%	estimated ± 5.2 pp, low confidence
98	Mistral Medium 3.5 128B	0.0%	estimated ± 3.1 pp, low confidence
99	Muse Glimmer 30B	0.0%	estimated ± 5.2 pp, low confidence
100	Nemotron 3 Super 100B	0.0%	estimated ± 5.0 pp, low confidence
101	Qwen3.5-122B-A10B	0.0%	estimated ± 5.2 pp, low confidence
102	Qwen3.5-27B	0.0%	estimated ± 3.1 pp, low confidence
103	Qwen3.5-35B-A3B	0.0%	estimated ± 3.1 pp, low confidence
104	Qwen3.5 397B	0.0%	estimated ± 3.1 pp, low confidence
105	Qwen3.6-27B	0.0%	estimated ± 3.1 pp, low confidence
106	Qwen3.6-35B-A3B	0.0%	estimated ± 3.1 pp, low confidence
107	Qwen3.6 Plus	0.0%	estimated ± 3.1 pp, low confidence
108	Qwen3.7 Plus	0.0%	estimated ± 5.0 pp, low confidence
109	Qwen3 Max	0.0%	estimated ± 3.1 pp, low confidence
110	Step 3.7 Flash	0.0%	estimated ± 3.1 pp, low confidence
111	Trinity-Large-Thinking	0.0%	estimated ± 3.1 pp, low confidence
112	UI-Mate-9B	0.0%	estimated ± 5.2 pp, low confidence

ApprenticeBench leaderboard

All leaderboards

Agentic · terminal

Agentic · tools

Coding

Math

Knowledge & reasoning

Instruction following

Multilingual

Vision & documents

Long context

General