模型测评：Artificial Analysis 指标

公开 LLM 指标聚合展示，包含智能、代码、数学、速度、价格、延迟和性价比等维度。

智能

Anthropic Claude Fable 5 (Adaptive Reasoning, Max Effort, Opus 4.8 Fallback)、Anthropic Claude Opus 4.8 (Adaptive Reasoning, Max Effort)、OpenAI GPT-5.5 (xhigh) 在 Artificial Analysis Intelligence Index 中排名靠前。

输出速度

Inception Mercury 2、IBM Granite 3.3 8B (Non-reasoning)、IBM Granite 4.0 H Small 的输出速度更高，适合强调响应速度的场景。

代码

Anthropic Claude Fable 5 (Adaptive Reasoning, Max Effort, Opus 4.8 Fallback)、OpenAI GPT-5.5 (xhigh)、OpenAI GPT-5.5 (high) 在代码能力指数中排名靠前。

数学

OpenAI GPT-5.2 (xhigh)、OpenAI GPT-5 Codex (high)、Google Gemini 3 Flash Preview (Reasoning) 在数学能力指数中排名靠前。

智能

Artificial Analysis Intelligence Index · 越高越好

Anthropic Claude Fable 5 (Adaptive Reasoning, Max Effort, Opus 4.8 Fallback)

Top 64.9平均 58.5310 项

代码

Artificial Analysis Coding Index · 越高越好

Anthropic Claude Fable 5 (Adaptive Reasoning, Max Effort, Opus 4.8 Fallback)

Top 62平均 56.3910 项

数学

Artificial Analysis Math Index · 越高越好

OpenAI GPT-5.2 (xhigh)

Top 99平均 96.5510 项

速度

输出 tokens/s · 越高越好

Inception Mercury 2

Top 1096.3平均 416.8610 项

各维度 Top 3 差距

每个维度以第 1 名为 100，对比第 2、3 名与头名的接近程度。

智能

#1Anthropic Claude Fable 5 (Adaptive Reasoning, Max Effort, Opus 4.8 Fallback)64.9

#2Anthropic Claude Opus 4.8 (Adaptive Reasoning, Max Effort)61.4

#3OpenAI GPT-5.5 (xhigh)60.2

代码

#1Anthropic Claude Fable 5 (Adaptive Reasoning, Max Effort, Opus 4.8 Fallback)62

#2OpenAI GPT-5.5 (xhigh)59.1

#3OpenAI GPT-5.5 (high)58.5

数学

#1OpenAI GPT-5.2 (xhigh)99

#2OpenAI GPT-5 Codex (high)98.7

#3Google Gemini 3 Flash Preview (Reasoning)97

速度

#1Inception Mercury 21096.3

#2IBM Granite 3.3 8B (Non-reasoning)407

#3IBM Granite 4.0 H Small406.2

混合价格

#1Alibaba Qwen3.5 0.8B (Reasoning)0.02

#2Alibaba Qwen3.5 0.8B (Non-reasoning)0.02

#3Google Gemma 3n E4B Instruct0.025

输入价格

#1Alibaba Qwen3.5 0.8B (Reasoning)0.01

#2Alibaba Qwen3.5 0.8B (Non-reasoning)0.01

#3Alibaba Qwen3.5 2B (Reasoning)0.02

输出价格

#1Google Gemma 3n E4B Instruct0.04

#2Alibaba Qwen3.5 0.8B (Reasoning)0.05

#3Alibaba Qwen3.5 0.8B (Non-reasoning)0.05

首响延迟

#1Cohere Command A+0.18

#2Cohere North Mini Code0.2

#3NVIDIA Nemotron Nano 9B V2 (Reasoning)0.23

性价比

#1Alibaba Qwen3.5 0.8B (Reasoning)525

#2Alibaba Qwen3.5 0.8B (Non-reasoning)495

#3Alibaba Qwen3.5 4B (Reasoning)451.7

供应商覆盖

#1Alibaba81

#2OpenAI67

#3Google55

智能 Top 5

直接对比头部 5 个模型的实际分数，比占比图更适合看排名差距。

#1Anthropic Claude Fable 5 (Adaptive Reasoning, Max Effort, Opus 4.8 Fallback)64.9

#2Anthropic Claude Opus 4.8 (Adaptive Reasoning, Max Effort)61.4

#3OpenAI GPT-5.5 (xhigh)60.2

#4OpenAI GPT-5.5 (high)58.9

#5Anthropic Claude Opus 4.7 (Adaptive Reasoning, Max Effort)57.3

智能

Artificial Analysis Intelligence Index · 越高越好

#1Anthropic Claude Fable 5 (Adaptive Reasoning, Max Effort, Opus 4.8 Fallback)64.9

#2Anthropic Claude Opus 4.8 (Adaptive Reasoning, Max Effort)61.4

#3OpenAI GPT-5.5 (xhigh)60.2

#4OpenAI GPT-5.5 (high)58.9

#5Anthropic Claude Opus 4.7 (Adaptive Reasoning, Max Effort)57.3

#6Google Gemini 3.1 Pro Preview57.2

#7OpenAI GPT-5.4 (xhigh)56.8

#8OpenAI GPT-5.5 (medium)56.7

#9Alibaba Qwen3.7 Max56.6

#10Google Gemini 3.5 Flash (high)55.3

显示前 10 项，完整模型名见图表下方列表。

代码

Artificial Analysis Coding Index · 越高越好

#1Anthropic Claude Fable 5 (Adaptive Reasoning, Max Effort, Opus 4.8 Fallback)62

#2OpenAI GPT-5.5 (xhigh)59.1

#3OpenAI GPT-5.5 (high)58.5

#4OpenAI GPT-5.4 (xhigh)57.2

#5Anthropic Claude Opus 4.8 (Adaptive Reasoning, Max Effort)56.7

#6OpenAI GPT-5.5 (medium)56.2

#7Google Gemini 3.1 Pro Preview55.5

#8OpenAI GPT-5.3 Codex (xhigh)53.1

#9Anthropic Claude Opus 4.7 (Non-reasoning, High Effort)53.1

#10Anthropic Claude Opus 4.7 (Adaptive Reasoning, Max Effort)52.5

显示前 10 项，完整模型名见图表下方列表。

数学

Artificial Analysis Math Index · 越高越好

#1OpenAI GPT-5.2 (xhigh)99

#2OpenAI GPT-5 Codex (high)98.7

#3Google Gemini 3 Flash Preview (Reasoning)97

#4OpenAI GPT-5.2 (medium)96.7

#5DeepSeek V3.2 Speciale96.7

#6Xiaomi MiMo-V2-Flash (Reasoning)96.3

#7OpenAI GPT-5.1 Codex (high)95.7

#8Google Gemini 3 Pro Preview (high)95.7

#9Z AI GLM-4.7 (Reasoning)95

#10KwaiKAT KAT-Coder-Pro V194.7

显示前 10 项，完整模型名见图表下方列表。

速度

输出 tokens/s · 越高越好

#1Inception Mercury 21096.3

#2IBM Granite 3.3 8B (Non-reasoning)407

#3IBM Granite 4.0 H Small406.2

#4OpenAI gpt-oss-120b (low)372.9

#5OpenAI gpt-oss-120b (high)359.3

#6Alibaba Qwen3.5 2B (Non-reasoning)351.3

#7NVIDIA Nemotron 3 Nano Omni 30B A3B Reasoning300

#8Google Gemini 2.5 Flash-Lite (Reasoning)293.4

#9NVIDIA Nemotron Nano 12B v2 VL (Reasoning)291.5

#10Google Gemini 3.1 Flash-Lite290.7

显示前 10 项，完整模型名见图表下方列表。

混合价格

每 100 万 tokens 混合价格 · 越低越好

#1Alibaba Qwen3.5 0.8B (Reasoning)0.02

#2Alibaba Qwen3.5 0.8B (Non-reasoning)0.02

#3Google Gemma 3n E4B Instruct0.025

#4Alibaba Qwen3.5 2B (Reasoning)0.04

#5Alibaba Qwen3.5 2B (Non-reasoning)0.04

#6Sarvam 30B (high)0.047

#7Meta Llama 3.2 Instruct 1B0.05

#8Google Gemma 3 4B Instruct0.05

#9Liquid AI LFM2 24B A2B0.052

#10Alibaba Qwen3.5 4B (Reasoning)0.06

显示前 10 项，完整模型名见图表下方列表。

输入价格

每 100 万输入 tokens · 越低越好

#1Alibaba Qwen3.5 0.8B (Reasoning)0.01

#2Alibaba Qwen3.5 0.8B (Non-reasoning)0.01

#3Alibaba Qwen3.5 2B (Reasoning)0.02

#4Alibaba Qwen3.5 2B (Non-reasoning)0.02

#5Google Gemma 3n E4B Instruct0.02

#6Sarvam 30B (high)0.026

#7Liquid AI LFM2 24B A2B0.03

#8Alibaba Qwen3.5 4B (Reasoning)0.03

#9Alibaba Qwen3.5 4B (Non-reasoning)0.03

#10IBM Granite 3.3 8B (Non-reasoning)0.03

显示前 10 项，完整模型名见图表下方列表。

输出价格

每 100 万输出 tokens · 越低越好

#1Google Gemma 3n E4B Instruct0.04

#2Alibaba Qwen3.5 0.8B (Reasoning)0.05

#3Alibaba Qwen3.5 0.8B (Non-reasoning)0.05

#4Meta Llama 3.2 Instruct 1B0.05

#5Google Gemma 3 4B Instruct0.08

#6Mistral Ministral 3 3B0.1

#7IBM Granite 4.1 8B0.1

#8Alibaba Qwen3.5 2B (Reasoning)0.1

#9Alibaba Qwen3.5 2B (Non-reasoning)0.1

#10Meta Llama 3.1 Instruct 8B0.1

显示前 10 项，完整模型名见图表下方列表。

首响延迟

Median time to first token · 越低越好

#1Cohere Command A+0.18

#2Cohere North Mini Code0.2

#3NVIDIA Nemotron Nano 9B V2 (Reasoning)0.23

#4Alibaba Qwen3.5 4B (Reasoning)0.23

#5Alibaba Qwen3.5 2B (Non-reasoning)0.24

#6Alibaba Qwen3.5 4B (Non-reasoning)0.26

#7NVIDIA Nemotron Nano 12B v2 VL (Reasoning)0.27

#8Cohere Command A0.28

#9Alibaba Qwen3.5 0.8B (Non-reasoning)0.28

#10NVIDIA Llama Nemotron Super 49B v1.5 (Non-reasoning)0.3

显示前 10 项，完整模型名见图表下方列表。

性价比

智能指数 / 混合价格 · 越高越好

#1Alibaba Qwen3.5 0.8B (Reasoning)525

#2Alibaba Qwen3.5 0.8B (Non-reasoning)495

#3Alibaba Qwen3.5 4B (Reasoning)451.7

#4Alibaba Qwen3.5 2B (Reasoning)407.5

#5Alibaba Qwen3.5 4B (Non-reasoning)376.7

#6Alibaba Qwen3.5 2B (Non-reasoning)367.5

#7Alibaba Qwen3.5 9B (Reasoning)286.7

#8Xiaomi MiMo-V2.5280

#9OpenAI gpt-oss-20B (high)278.4

#10Xiaomi MiMo-V2-Flash (Feb 2026)276.7

显示前 10 项，完整模型名见图表下方列表。

供应商覆盖

当前 API 返回的模型数量 · 越高代表覆盖更多

#1Alibaba81

#2OpenAI67

#3Google55

#4Anthropic32

#5Mistral32

#6DeepSeek31

#7xAI20

#8NVIDIA18

#9Z AI18

#10Meta17

显示前 10 项，完整模型名见图表下方列表。