Agentic

Agentic

How well models plan, use tools, and complete multi-step tasks.

Last updated · May 31, 2026, 13:27 UTC

RankModelCompositeTrend
1
Qwen3-Maxlimited
Alibaba
82.2
2
Nemotron-Orchestrator-8Blimited
NVIDIA
76.3
3
Gemini 3 Prolimited
Google
74.3
4
DeepSeek V3.2limited
DeepSeek
71.2
5
Claude Sonnet 4.5
Anthropic
71.0
6
Claude Opus 4.1
Anthropic
68.8
7
Qwen3.5limited
Alibaba
68.4
8
Claude Opus 4
Anthropic
67.5
9
DeepSeek V4 Prolimited
DeepSeek
67.2
10
GLM-5.1limited
Z.ai
67.1
11
GPT-5.2limited
OpenAI
66.6
12
Claude Opus 4.5limited
Anthropic
66.1
13
GPT-5
OpenAI
64.6
14
GLM-5limited
Z.ai
63.2
15
Gemini 3 Flashlimited
Google
61.7
16
Claude Sonnet 4
Anthropic
60.3
17
Kimi K2limited
Moonshot AI
58.8
18
Claude 3.7 Sonnet
Anthropic
58.2
19
Claude Haiku 4.5limited
Anthropic
56.4
20
MiniMax M2.5limited
MiniMax
55.6
21
GPT-4.1-minilimited
OpenAI
53.0
22
GPT-5 Codexlimited
OpenAI
52.7
23
GPT-5.1limited
OpenAI
51.3
24
o4-mini
OpenAI
50.8
25
GPT-4.1
OpenAI
47.6
26
o3
OpenAI
44.7
27
Grok 4limited
xAI
41.5
28
Claude Opus 4.6limited
Anthropic
38.8
29
GPT-5.5limited
OpenAI
37.4
30
Doubao-Seed-Codelimited
ByteDance
36.5
31
o3-minilimited
OpenAI
32.3
32
GPT-5.4limited
OpenAI
30.7
33
DeepSeek R1
DeepSeek
30.1
34
DeepSeek V3
DeepSeek
26.9
35
GPT-4olimited
OpenAI
26.8
36
Claude Opus 4.7limited
Anthropic
25.3
37
Grok 3limited
xAI
24.6
38
Gemini 3.1 Pro Previewlimited
Google
22.5
39
Gemini 2.5 Prolimited
Google
19.4
40
Gemini 2.5 Flashlimited
Google
18.8