Anthropic

Claude Code：Coding Agent Index 得分、性能与价格

与其他编程智能体相比，Claude Code 的 Coding Agent Index 得分处于领先行列，但价格尤其昂贵。此外，其速度明显较慢。其最佳结果使用 Claude Opus 5.5 (Max, Default Fallback)，我们共测试了 6 个模型。

Claude Code 搭配 Claude Opus 5.5 (Max, Default Fallback) 在 Artificial Analysis Coding Agent Index 上的得分为 62，在编程智能体中远高于平均水平（中位数：49）。

Claude Code 搭配 Claude Opus 5.5 (Max, Default Fallback) 每项任务耗时 1.0h，明显较慢（中位数：36.1m）。

在 Coding Agent Index 上运行 Claude Code 搭配 Claude Opus 5.5 (Max, Default Fallback) 的平均每项任务 API 成本为 $11.42（昂贵，中位数：$5.41）。

性能

Artificial Analysis Coding Agent Index上的性能。

Artificial Analysis Coding Agent Index

Artificial Analysis Coding Agent Index v2.0 incorporates 3 benchmarks: FrontierCode v1.1, Terminal-Bench 4.0, and SWE-Atlas-QnA-AA · Higher is better

着色方式

模型路由器

The Artificial Analysis Coding Agent Index is a composite score built from FrontierCode v1.1、Terminal-Bench 4.0和SWE-Atlas-QnA-AA.

It is useful for quick comparison, but it should be read alongside the per-eval breakdowns. Two agents with similar index values can still have different strengths across repository tasks, terminal workflows, and rubric-based evaluations.

Artificial Analysis Coding Agent Index vs. 每个任务的成本

Artificial Analysis Coding Agent Index vs. 每个任务的平均按 token 计费 API 成本（USD）

着色方式

模型路由器

Most attractive quadrant

Pareto line

Each point represents a coding-agent variant. Farther left means lower average cost per task, while higher on the chart means higher benchmark performance. The most efficient agents sit toward the upper-left: stronger results at lower cost.

Token 用量

Artificial Analysis Coding Agent Index上的 token 消耗，包括总用量、token 构成、效率以及各基准测试的细分。

每个任务的 Token 用量

每个任务的平均输入、缓存和输出 token

模型路由器

提示词缓存命中率可能因提供商路由而显著变化，从而实质性地影响实际成本。

Non-cached input tokens sent to the model, including prompts, instructions, tool context, and task context that were not served from prompt cache.

Artificial Analysis Coding Agent Index vs. 总 Token

Artificial Analysis Coding Agent Index vs. 每个任务的平均总 Token

着色方式

模型路由器

Most attractive quadrant

Each point represents a coding-agent variant. Farther left means lower average total token usage per task, while higher on the chart means higher benchmark performance. Agents toward the upper-left achieve stronger results with fewer tokens.

成本

基于当前按 token 计费的 API 价格的 Artificial Analysis Coding Agent Index成本，在可用时包括缓存写入价格和缓存折扣。许多用户通过订阅套餐而非按 token 付费来使用编程智能体的 harness。

每个任务的成本

Average pay-per-token API cost per task (USD) · Lower is better

着色方式

模型路由器

This chart shows the average pay-per-token API cost per task across the Artificial Analysis Coding Agent Index, spanning FrontierCode v1.1、Terminal-Bench 4.0和SWE-Atlas-QnA-AA.

Where applicable, that cost model includes standard input pricing, discounted cached-input pricing, separate cache-write charges, and output pricing rather than treating all prompt tokens as if they were billed at the same uncached input rate.

It is intended to show pay-per-token API cost, not consumer plan pricing or the full operational cost of deploying the system in production. Infrastructure, engineering, and supervision costs are not the focus of this metric.

执行时间

Artificial Analysis Coding Agent Index上的智能体活跃运行时间。

每个任务的时间

Average agent wall time per task · Lower is better

着色方式

模型路由器

This chart uses agent wall time: how long the agent process was actively running on each task.

It does not include environment startup, verifier or judge time, or other harness overhead, so it is a cleaner comparison of how long the agent itself was working.

Artificial Analysis Coding Agent Index vs. 执行时间

Artificial Analysis Coding Agent Index vs. 每个任务的平均智能体运行时间

着色方式

模型路由器

Most attractive quadrant

Each point represents a coding-agent variant. Farther left means shorter average agent runtime per task, while higher on the chart means higher benchmark performance. Agents toward the upper-left deliver stronger results in less active agent time.

安全拒绝

方法说明

Artificial Analysis Coding Agent Index 中提供商或模型出于安全原因拒绝继续任务的频率。

安全拒绝率

Share of retained task attempts with a safety refusal · Equal weight per benchmark · Lower is better

分组依据

模型路由器

安全与回退行为由提供商配置，可能随时间变化；这些数值反映的是基准测试时记录的行为。

受阻指尝试就此结束。回退指由另一个模型完成该次尝试。两者都计入拒绝比例。

奖励作弊

方法说明

原本能够通过、但因完整性违规而被计为零分的尝试，占全部任务尝试的比例。

奖励作弊率：FrontierCode v1.1

Share of task attempts scored zero for reward hacking · 使用 Cognition 规则扫描器的修改版检测对解答材料的违规访问 · Lower is better

着色方式

模型路由器

Claude Code：Coding Agent Index 得分、性能与价格

比较摘要

性能

Artificial Analysis Coding Agent Index

What This Metric Means

Artificial Analysis Coding Agent Index vs. 每个任务的成本

How to Read This Chart

Token 用量

每个任务的 Token 用量

Input Tokens

Artificial Analysis Coding Agent Index vs. 总 Token

How to Read This Chart

成本

每个任务的成本

What Cost Is Measuring

执行时间

每个任务的时间

What Execution Time Is Measuring

Artificial Analysis Coding Agent Index vs. 执行时间

How to Read This Chart

安全拒绝

安全拒绝率

安全拒绝的结果

奖励作弊

奖励作弊率：FrontierCode v1.1

运行规格