The Agents Leaderboard ranks large language models by their ability to act as autonomous problem-solvers rather than simply generate responses. Unlike standard text-generation evaluations, agentic benchmarks test whether a model can independently break down complex goals, execute shell commands in terminal environments, invoke external APIs, interpret tool outputs, and recover from runtime errors.

Drawing on evaluations agent-focused benchmarks this leaderboard highlights models capable of handling complex, multi-step tasks across developer and enterprise workflows.

Looking for well-rounded performance? Agentic capabilities account for 20% of the composite evaluation. Check the Overall LLM Ranking to see how top agentic models perform across coding, mathematics, and complex reasoning.

Last updated: October 19, 2026
Rank Model Name Relative Quality Score
1 Claude Opus 5.5
Quality of Agents 100.00%

100.00
2 GPT-6 Astra
Quality of Agents 93.90%

93.90
3 Muse Wonder Full
Quality of Agents 92.17%

92.17
4 Claude Sonnet 5.5
Quality of Agents 89.41%

89.41
5 Claude Fable 5.1
Quality of Agents 88.90%

88.90
6 GPT-6.1 Sol
Quality of Agents 85.30%

85.30
7 Gemini 4 Argon
Quality of Agents 84.40%

84.40
8 GPT-6 Sol
Quality of Agents 77.32%

77.32
9 Claude Opus 5
Quality of Agents 76.23%

76.23
10 GPT-5.5 Pro
Quality of Agents 74.96%

74.96
11 Claude Opus 4.7
Quality of Agents 73.74%

73.74
12 Claude Fable 5
Quality of Agents 73.45%

73.45
13 Gemini 3.5 Flash
Quality of Agents 70.43%

70.43
14 GPT-5.6 Sol
Quality of Agents 70.27%

70.27
15 Ling 3.0 Flash Fin
Quality of Agents 70.09%

70.09
16 DeepSeek-V4.1-Flash
Quality of Agents 69.51%

69.51
17 Claude Opus 4.6
Quality of Agents 68.70%

68.70
18 Muse Wonder
Quality of Agents 68.54%

68.54
19 Claude Opus 4.8
Quality of Agents 68.32%

68.32
20 Hy4
Quality of Agents 67.26%

67.26
21 Kimi K2.6
Quality of Agents 67.13%

67.13
22 Gemini 3.6 Flash
Quality of Agents 66.78%

66.78
23 Kimi K2.7 Code
Quality of Agents 66.43%

66.43
24 MiMo-V2.6-Pro
Quality of Agents 65.71%

65.71
25 Claude Haiku 5.5
Quality of Agents 65.65%

65.65
26 MiMo-V2.6-Flash
Quality of Agents 65.07%

65.07
27 Grok 4.7
Quality of Agents 64.90%

64.90
28 DeepSeek-V4-Pro
Quality of Agents 64.24%

64.24
29 Muse Spark 1.3
Quality of Agents 63.26%

63.26
30 Kimi K3
Quality of Agents 62.79%

62.79
31 GPT-5.6 Terra
Quality of Agents 62.53%

62.53
32 GLM-5.1
Quality of Agents 62.43%

62.43
33 GLM-5.2
Quality of Agents 62.06%

62.06
34 GPT-6 Luna
Quality of Agents 61.63%

61.63
35 Qwen3.8
Quality of Agents 61.50%

61.50
36 GPT-5.5
Quality of Agents 61.24%

61.24
37 Qwen3.8 Flash
Quality of Agents 60.87%

60.87
38 GPT-5.3 Codex
Quality of Agents 60.87%

60.87
39 MiniMax M3.1 Flash
Quality of Agents 60.14%

60.14
40 DeepSeek-V4-Flash
Quality of Agents 59.81%

59.81
41 GLM-5.3
Quality of Agents 59.50%

59.50
42 Claude Sonnet 5
Quality of Agents 59.46%

59.46
43 Grok 4.6
Quality of Agents 59.26%

59.26
44 GPT-5.4
Quality of Agents 58.80%

58.80
45 Qwen3.8-Flash-Next
Quality of Agents 58.19%

58.19
46 Gemini 3.8 Flash
Quality of Agents 57.91%

57.91
47 MiMo-V2.5
Quality of Agents 57.74%

57.74
48 Qwen3.6 Plus
Quality of Agents 56.70%

56.70
49 Qwen3.5-397B-A17B
Quality of Agents 55.83%

55.83
50 GLM-5
Quality of Agents 55.13%

55.13
51 GPT-5.4 mini
Quality of Agents 54.43%

54.43
52 Ling 3.0 Flash
Quality of Agents 54.43%

54.43
53 GPT-5.2 Codex
Quality of Agents 54.43%

54.43
54 GLM-5V-Turbo
Quality of Agents 53.57%

53.57
55 Grok 4.5
Quality of Agents 53.28%

53.28
56 Muse Spark 1.1
Quality of Agents 53.21%

53.21
57 GLM-5.3-Flash
Quality of Agents 52.87%

52.87
58 Kimi K2.5
Quality of Agents 52.35%

52.35
59 GPT-5.2 Pro
Quality of Agents 52.35%

52.35
60 MiniMax M2.1
Quality of Agents 52.00%

52.00
61 Claude Opus 4.5
Quality of Agents 51.65%

51.65
62 Gemini 3 Flash
Quality of Agents 51.65%

51.65
63 Gemini 3.7 Flash
Quality of Agents 51.55%

51.55
64 MiniMax M2.5
Quality of Agents 51.30%

51.30
65 Claude Sonnet 4.5
Quality of Agents 50.78%

50.78
66 Seed 2.0 Pro
Quality of Agents 50.61%

50.61
67 Qwen3.7
Quality of Agents 50.36%

50.36
68 GPT-5.6 Luna
Quality of Agents 50.35%

50.35
69 MiMo-V2-Pro
Quality of Agents 50.26%

50.26
70 Qwen3.6-27B
Quality of Agents 49.04%

49.04
71 Hy3
Quality of Agents 48.80%

48.80
72 Kimi K2
Quality of Agents 48.52%

48.52
73 MiniMax M2.7
Quality of Agents 48.35%

48.35
74 Muse Spark Open
Quality of Agents 48.25%

48.25
75 Qwen3.5-122B-A10B
Quality of Agents 48.17%

48.17
76 Gemini 3.5 Flash-Lite
Quality of Agents 48.17%

48.17
77 Step-3.5-Flash
Quality of Agents 47.83%

47.83
78 Qwen3.8-27B
Quality of Agents 47.75%

47.75
79 Gemini 3 Pro
Quality of Agents 47.65%

47.65
80 Muse Spark
Quality of Agents 46.61%

46.61
81 Qwen3.5-27B
Quality of Agents 46.61%

46.61
82 GPT-5.2
Quality of Agents 46.43%

46.43
83 Qwen3.7-Plus
Quality of Agents 46.25%

46.25
84 Claude Sonnet 4.6
Quality of Agents 44.72%

44.72
85 Step 3.7 Flash
Quality of Agents 44.35%

44.35
86 MiMo-V2.5-Pro
Quality of Agents 44.02%

44.02
87 Qwen3 VL 235B A22B
Quality of Agents 43.83%

43.83
88 LongCat-Flash
Quality of Agents 43.48%

43.48
89 Qwen3 VL 32B
Quality of Agents 43.48%

43.48
90 Mistral Large 4
Quality of Agents 43.44%

43.44
91 Claude Opus 4.1
Quality of Agents 43.30%

43.30
92 Gemini 3.1 Pro
Quality of Agents 43.25%

43.25
93 Gemma 4 26B-A4B
Quality of Agents 43.13%

43.13
94 Qwen3.5-35B-A3B
Quality of Agents 43.13%

43.13
95 Qwen3-Next-80B-A3B
Quality of Agents 42.26%

42.26
96 DeepSeek-V3.2
Quality of Agents 41.91%

41.91
97 Qwen3.6-35B-A3B
Quality of Agents 41.57%

41.57
98 GPT-5.4 nano
Quality of Agents 40.87%

40.87
99 Qwen3 235B A22B
Quality of Agents 40.87%

40.87
100 DeepSeek-V3.2-Speciale
Quality of Agents 40.00%

40.00