Agentic · tools

DRACO leaderboard

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

As of 2026-10-07, the highest measured score on DRACO is 88.6% by Claude Opus 5. 117 more models have estimated scores, calibrated from the benchmarks they were measured on.

Measured scores: benchlm.ai.

#	Model	Score	Source
1	Claude Opus 5.5	89.6%	estimated ± 3.0 pp, low confidence
2	Claude Opus 5	88.6%	measured
3	Claude Fable 5.1	87.3%	estimated ± 3.0 pp, medium confidence
4	Claude Sonnet 5.5	87.0%	measured
5	Grok 4.7	86.5%	estimated ± 3.0 pp, medium confidence
6	MiMo-V2.6-Pro	85.9%	estimated ± 3.0 pp, medium confidence
7	Muse Spark 1.3	85.8%	estimated ± 3.0 pp, medium confidence
8	Qwen3.8 Max Preview	85.6%	estimated ± 3.0 pp, medium confidence
9	Ling 3.1 Flash	85.5%	measured
10	GLM-5.3	85.2%	estimated ± 3.0 pp, medium confidence
11	Qwen3.8-Flash-Next	84.8%	estimated ± 3.0 pp, medium confidence
12	Gemini 4 Argon	84.6%	estimated ± 3.0 pp, medium confidence
13	Grok 4.6	84.6%	estimated ± 3.0 pp, medium confidence
14	Claude Fable 5	84.4%	estimated ± 3.0 pp, medium confidence
15	GPT-5.6 Sol	84.4%	estimated ± 3.0 pp, medium confidence
16	MiMo-V2.6-Flash	84.3%	estimated ± 3.0 pp, medium confidence
17	DeepSeek V4.1 Flash	84.1%	estimated ± 3.0 pp, medium confidence
18	DeepSeek V4 Pro 0813	83.9%	estimated ± 3.0 pp, medium confidence
19	GPT-6.1 Sol	83.6%	estimated ± 3.0 pp, medium confidence
20	Step 5 Preview	83.3%	measured
21	GPT-6 Astra	82.9%	estimated ± 3.0 pp, medium confidence
22	Kimi K3	82.8%	estimated ± 3.0 pp, medium confidence
23	GPT-6 Sol	82.3%	estimated ± 3.0 pp, medium confidence
24	Muse Spark 1.2	81.6%	estimated ± 3.0 pp, medium confidence
25	Claude Sonnet 5	81.4%	estimated ± 3.0 pp, medium confidence
26	GPT-5.6 Luna	81.3%	estimated ± 3.0 pp, medium confidence
27	Claude Opus 4.8	81.2%	estimated ± 3.0 pp, medium confidence
28	GPT-5.6 Terra	81.1%	estimated ± 3.0 pp, medium confidence
29	DeepSeek V4 Flash 0731	80.8%	estimated ± 3.0 pp, medium confidence
30	GPT-6 Luna	80.8%	estimated ± 3.0 pp, medium confidence
31	Gemini 3.8 Flash	80.7%	estimated ± 3.0 pp, medium confidence
32	Mistral Large 4	80.5%	estimated ± 3.0 pp, medium confidence
33	Qwen3.8-27B	80.5%	estimated ± 3.0 pp, medium confidence
34	Gemini 3.7 Flash	79.8%	estimated ± 3.0 pp, medium confidence
35	Grok 4.5	79.8%	estimated ± 3.0 pp, medium confidence
36	GLM-5.2	79.5%	estimated ± 3.0 pp, medium confidence
37	Claude Opus 4.7 (Adaptive)	79.1%	estimated ± 3.0 pp, medium confidence
38	GPT-5.5	79.1%	estimated ± 3.0 pp, medium confidence
39	Gemini 3.5 Flash	78.9%	estimated ± 3.0 pp, medium confidence
40	Gemini 3.6 Flash	77.7%	estimated ± 3.0 pp, medium confidence
41	Hy4 preview	77.2%	measured
42	GPT-5.4	76.9%	estimated ± 3.0 pp, medium confidence
43	MiniMax M3	76.8%	estimated ± 3.0 pp, medium confidence
44	Hy3 Preview	76.2%	estimated ± 3.0 pp, medium confidence
45	Muse Spark 1.1	76.2%	estimated ± 3.0 pp, medium confidence
46	Apodex 1.1	75.8%	estimated ± 3.0 pp, medium confidence
47	Apodex 1.1 Mini	75.8%	estimated ± 3.0 pp, medium confidence
48	Quasar 438B	75.6%	estimated ± 3.0 pp, medium confidence
49	Ling 3.0 Flash VL	75.2%	estimated ± 3.0 pp, medium confidence
50	Nemotron 3 Ultra	75.1%	estimated ± 3.0 pp, medium confidence
51	Qwen3.7 Max	74.5%	estimated ± 3.0 pp, medium confidence
52	Inkling-Small	74.3%	estimated ± 3.0 pp, medium confidence
53	MiMo-V2.5-Pro	74.3%	estimated ± 3.0 pp, medium confidence
54	GLM-5.1	74.3%	estimated ± 3.0 pp, medium confidence
55	Solar Pro 4	74.1%	estimated ± 3.0 pp, medium confidence
56	Grok 4.3	73.5%	estimated ± 3.0 pp, medium confidence
57	Inkling	73.4%	estimated ± 3.0 pp, medium confidence
58	Hy3	73.2%	estimated ± 3.0 pp, medium confidence
59	Kimi K2.6	72.6%	estimated ± 3.0 pp, medium confidence
60	Kimi K2.7 Code	72.6%	estimated ± 3.0 pp, medium confidence
61	GLM-4.7	72.1%	estimated ± 3.0 pp, medium confidence
62	GPT-5.4 mini	72.1%	estimated ± 3.0 pp, medium confidence
63	Step 3.7 Flash	72.1%	estimated ± 3.0 pp, medium confidence
64	MiniMax M2.7	72.1%	estimated ± 3.0 pp, medium confidence
65	Muse Spark	71.8%	estimated ± 3.0 pp, medium confidence
66	Qwen3.6 Plus	71.7%	estimated ± 3.0 pp, medium confidence
67	Qwen3.6-27B	71.6%	estimated ± 3.0 pp, medium confidence
68	Gemini 3.5 Flash-Lite	71.5%	estimated ± 3.0 pp, medium confidence
69	A.X K2	71.1%	estimated ± 3.0 pp, medium confidence
70	GPT-5.4 nano	70.8%	estimated ± 3.0 pp, medium confidence
71	Ling 3.0 Flash FP8	70.7%	estimated ± 3.0 pp, medium confidence
72	Ling 3.0 Flash	70.4%	measured
73	GPT-5 (high)	70.2%	estimated ± 3.0 pp, low confidence
74	Qwen3.6-35B-A3B	69.7%	estimated ± 3.0 pp, low confidence
75	Kimi K2.5	68.6%	estimated ± 3.0 pp, low confidence
76	Kimi K2.5 (Reasoning)	68.6%	estimated ± 3.0 pp, low confidence
77	GPT-5.1	68.3%	estimated ± 3.0 pp, low confidence
78	Qwen3.5-122B-A10B	68.0%	estimated ± 3.0 pp, low confidence
79	Gemini 3.1 Pro	67.6%	estimated ± 3.0 pp, low confidence
80	Muse Glimmer 30B	67.5%	estimated ± 3.0 pp, low confidence
81	Qwen3.7 Plus	67.1%	estimated ± 3.0 pp, low confidence
82	Mistral Medium 3.5 128B	67.0%	estimated ± 3.0 pp, low confidence
83	MiniCPM5-2B	65.9%	estimated ± 3.0 pp, low confidence
84	Nemotron 3.5 Lightning 30B A3B NVFP4	64.4%	estimated ± 3.0 pp, low confidence
85	MiMo-V2-Flash	64.3%	estimated ± 3.0 pp, low confidence
86	Gemma 4 31B	64.0%	estimated ± 3.0 pp, low confidence
87	GPT-OSS 120B	63.8%	estimated ± 3.0 pp, low confidence
88	Granite 4.2 30B	63.1%	estimated ± 3.0 pp, low confidence
89	Gemma 4 26B A4B	62.9%	estimated ± 3.0 pp, low confidence
90	Ling 3.0 Tiny	62.9%	estimated ± 3.0 pp, low confidence
91	Command A+	61.9%	estimated ± 3.0 pp, low confidence
92	Celeris-1	61.5%	estimated ± 3.0 pp, low confidence
93	DeepSeek V3	61.5%	estimated ± 3.0 pp, low confidence
94	DeepSeek V3 0324	61.5%	estimated ± 3.0 pp, low confidence
95	Gemini 2.5 Pro	61.5%	estimated ± 3.0 pp, low confidence
96	Gemma 3 27B	61.5%	estimated ± 3.0 pp, low confidence
97	Gemma 4 12B	61.5%	estimated ± 3.0 pp, low confidence
98	Gemma 4 E2B	61.5%	estimated ± 3.0 pp, low confidence
99	Gemma 4 E4B	61.5%	estimated ± 3.0 pp, low confidence
100	GPT-4.1 mini	61.5%	estimated ± 3.0 pp, low confidence
101	GPT-4.1 nano	61.5%	estimated ± 3.0 pp, low confidence
102	GPT-4o	61.5%	estimated ± 3.0 pp, low confidence
103	GPT-4o mini	61.5%	estimated ± 3.0 pp, low confidence
104	GPT-OSS 20B	61.5%	estimated ± 3.0 pp, low confidence
105	Granite 4.2 3B	61.5%	estimated ± 3.0 pp, low confidence
106	Granite 4.2 8B	61.5%	estimated ± 3.0 pp, low confidence
107	K-Exaone	61.5%	estimated ± 3.0 pp, low confidence
108	LFM2.5-2.6B	61.5%	estimated ± 3.0 pp, low confidence
109	Ling 2.6 Flash	61.5%	estimated ± 3.0 pp, low confidence
110	Llama 4 Maverick	61.5%	estimated ± 3.0 pp, low confidence
111	Llama 4 Scout	61.5%	estimated ± 3.0 pp, low confidence
112	Mercury 2.5	61.5%	estimated ± 3.0 pp, low confidence
113	Mistral Large 3	61.5%	estimated ± 3.0 pp, low confidence
114	Mistral Small 4	61.5%	estimated ± 3.0 pp, low confidence
115	Mistral Small 4 (Reasoning)	61.5%	estimated ± 3.0 pp, low confidence
116	Nemotron 3 Nano 30B	61.5%	estimated ± 3.0 pp, low confidence
117	Nemotron 3 Nano Omni 30B A3B	61.5%	estimated ± 3.0 pp, low confidence
118	Nemotron 3 Super 100B	61.5%	estimated ± 3.0 pp, low confidence
119	North Mini Code	61.5%	estimated ± 3.0 pp, low confidence
120	Solar Pro 3	61.5%	estimated ± 3.0 pp, low confidence
121	Trinity-Large-Preview	61.5%	estimated ± 3.0 pp, low confidence
122	Trinity-Large-Thinking	61.5%	estimated ± 3.0 pp, low confidence
123	Ultravox v0.6 Llama 3.3 70B	61.5%	estimated ± 3.0 pp, low confidence

DRACO leaderboard

All leaderboards

Agentic · terminal

Agentic · tools

Coding

Math

Knowledge & reasoning

Instruction following

Multilingual

Vision & documents

Long context

General