第 1 名
Claude Opus 5 (Adaptive Reasoning, Max Effort)
Anthropic
能力得分
55.3
54 tok/s
输出速度
全部天梯10 / 50 48 个模型
当前数据AIUtil 能力天梯
模型 Agent 能力天梯
聚焦规划、工具调用与多步骤任务执行能力。
本次数据
收录 50 个模型
当前第一
Claude Opus 5 (Adaptive Reasoning, Max Effort)
最高分
55.3
入榜数量
50
最近发布
2026/08/05
前三名
领先梯队
分数越高，排名越靠前
第 2 名
Claude Opus 5 (Adaptive Reasoning, Xhigh Effort)
Anthropic
能力得分
54.5
55 tok/s
输出速度
第 3 名
GPT-5.6 Sol (max)
OpenAI
能力得分
54.0
64 tok/s
输出速度
分数对比
前十名能力差距
- 01Claude Opus 5 (Adaptive Reasoning, Max Effort)55.3
- 02Claude Opus 5 (Adaptive Reasoning, Xhigh Effort)54.5
- 03GPT-5.6 Sol (max)54.0
- 04Claude Fable 5 (Adaptive Reasoning, Max Effort, Opus 4.8 Fallback)52.8
- 05Claude Opus 5 (Adaptive Reasoning, High Effort)52.1
- 06GPT-5.6 Sol (xhigh)51.8
- 07Kimi K3 (max)50.1
- 08GPT-5.6 Sol (high)48.5
- 09GPT-5.6 Terra (max)47.4
- 10Claude Opus 4.8 (Adaptive Reasoning, Max Effort)47.2
015304560Artificial Analysis Agentic Index
价格对比
能力与输出价格
越靠近左上角，代表更高能力与更低输出价格。
每百万输出 Token 价格（美元，对数轴）Artificial Analysis Agentic Index
58.350.643.035.327.6$0.28$1.02$3.74$14$50
Claude Opus 5 (Adaptive Reasoning, Max Effort)
Anthropic
能力分55.3输出价格$25 / 1M
54 tok/s全部结果
完整排名
Artificial Analysis Agentic Index
| 排名 | 模型 | 厂商 | Artificial Analysis Agentic Index | 输出速度 | 发布日期 |
|---|---|---|---|---|---|
| #01 | Claude Opus 5 (Adaptive Reasoning, Max Effort) | Anthropic | 55.3 | 54 tok/s | |
| #02 | Claude Opus 5 (Adaptive Reasoning, Xhigh Effort) | Anthropic | 54.5 | 55 tok/s | |
| #03 | GPT-5.6 Sol (max) | OpenAI | 54.0 | 64 tok/s | |
| #04 | Claude Fable 5 (Adaptive Reasoning, Max Effort, Opus 4.8 Fallback) | Anthropic | 52.8 | 72 tok/s | |
| #05 | Claude Opus 5 (Adaptive Reasoning, High Effort) | Anthropic | 52.1 | 53 tok/s | |
| #06 | GPT-5.6 Sol (xhigh) | OpenAI | 51.8 | 63 tok/s | |
| #07 | Kimi K3 (max) | Kimi | 50.1 | 38 tok/s | |
| #08 | GPT-5.6 Sol (high) | OpenAI | 48.5 | 61 tok/s | |
| #09 | GPT-5.6 Terra (max) | OpenAI | 47.4 | 125 tok/s | |
| #10 | Claude Opus 4.8 (Adaptive Reasoning, Max Effort) | Anthropic | 47.2 | 58 tok/s | |
| #11 | Claude Opus 5 (Adaptive Reasoning, Medium Effort) | Anthropic | 47.1 | 54 tok/s | |
| #12 | Muse Spark 1.2 (xhigh) | Meta | 46.8 | 0 tok/s | |
| #13 | Claude Sonnet 5 (Adaptive Reasoning, Max Effort) | Anthropic | 46.7 | 70 tok/s | |
| #14 | DeepSeek V4 Flash 0731 (Reasoning, Max Effort) | DeepSeek | 45.7 | 103 tok/s | |
| #15 | Grok 4.5 (high) | SpaceXAI | 45.7 | 59 tok/s | |
| #16 | GPT-5.6 Luna (max) | OpenAI | 45.6 | 178 tok/s | |
| #17 | GPT-5.5 (xhigh) | OpenAI | 44.9 | 76 tok/s | |
| #18 | GPT-5.6 Terra (xhigh) | OpenAI | 44.7 | 116 tok/s | |
| #19 | GPT-5.6 Sol (medium) | OpenAI | 44.5 | 58 tok/s | |
| #20 | Claude Opus 4.7 (Adaptive Reasoning, Max Effort) | Anthropic | 44.4 | 51 tok/s | |
| #21 | GPT-5.5 (high) | OpenAI | 43.5 | 72 tok/s | |
| #22 | GLM-5.2 (max) | Z AI | 43.1 | 150 tok/s | |
| #23 | GPT-5.6 Luna (xhigh) | OpenAI | 42.9 | 164 tok/s | |
| #24 | GPT-5.6 Terra (high) | OpenAI | 41.3 | 109 tok/s | |
| #25 | GPT-5.4 (xhigh) | OpenAI | 41.1 | 125 tok/s | |
| #26 | Claude Sonnet 4.6 (Adaptive Reasoning, Max Effort) | Anthropic | 40.8 | 58 tok/s | |
| #27 | GPT-5.6 Luna (high) | OpenAI | 40.1 | 163 tok/s | |
| #28 | GPT-5.6 Sol (low) | OpenAI | 40.0 | 61 tok/s | |
| #29 | Claude Opus 5 (Adaptive Reasoning, Low Effort) | Anthropic | 39.8 | 52 tok/s | |
| #30 | Gemini 3.6 Flash (high) | 38.7 | 188 tok/s | ||
| #31 | GPT-5.5 (medium) | OpenAI | 37.8 | 69 tok/s | |
| #32 | Muse Spark 1.1 (xhigh) | Meta | 37.5 | 213 tok/s | |
| #33 | Gemini 3.5 Flash (high) | 37.4 | 168 tok/s | ||
| #34 | GPT-5.6 Terra (medium) | OpenAI | 37.0 | 112 tok/s | |
| #35 | Kimi K3 (low) | Kimi | 37.0 | 38 tok/s | |
| #36 | DeepSeek V4 Pro (Reasoning, Max Effort) | DeepSeek | 36.4 | 63 tok/s | |
| #37 | MiniMax-M3 | MiniMax | 35.4 | 82 tok/s | |
| #38 | GPT-5.6 Sol (Non-reasoning) | OpenAI | 34.9 | 60 tok/s | |
| #39 | JT-4.1 Flash 236B A21B | China Mobile | 34.9 | 0 tok/s | |
| #40 | Motif 3 (Beta) | Motif Technologies | 34.9 | 0 tok/s | |
| #41 | GLM-5.2 (Non-reasoning) | Z AI | 34.8 | 97 tok/s | |
| #42 | DeepSeek V4 Pro (Reasoning, High Effort) | DeepSeek | 34.4 | 64 tok/s | |
| #43 | Claude Sonnet 5 (Non-reasoning, High Effort) | Anthropic | 33.7 | 62 tok/s | |
| #44 | Inkling (xhigh) | Thinking Machines | 32.3 | 81 tok/s | |
| #45 | DeepSeek V4 Flash (Reasoning, Max Effort) | DeepSeek | 31.1 | 0 tok/s | |
| #46 | GPT-5.6 Luna (medium) | OpenAI | 31.0 | 151 tok/s | |
| #47 | Nex-N2-Pro | Nex AGI | 31.0 | 129 tok/s | |
| #48 | Inkling Small | Thinking Machines | 30.8 | 136 tok/s | |
| #49 | Hy3 | Tencent | 30.7 | 65 tok/s | |
| #50 | GPT-5.6 Terra (low) | OpenAI | 30.6 | 106 tok/s |