Codex vs. Kimi Code CLI

Artificial Analysis Coding Agent Index의 벤치마크 점수, 비용, 실행 시간, 토큰 사용량을 기준으로 Codex와 Kimi Code CLI를 비교합니다.

방법론에 관한 자세한 내용은 방법론 페이지를 참조하세요.

다른 비교 살펴보기
vs

주요 내용

Coding Agent Index

Artificial Analysis Coding Agent Index v1.3 · Higher is better

Time per Task

Average agent wall time per task · Lower is better

Cost per Task

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

Comparison

Side-by-side comparison of Codex and Kimi Code CLI.

Coding Agent Comparison

Metric
Codex
GPT-5.6 Sol (max)
Kimi Code CLI
Kimi K3
Analysis
Agent Harness
Codex
Kimi Code CLI
Representative Model
GPT-5.6 Sol (max)
Kimi K3
Coding Agent Index
67
61
Codex has a higher Coding Agent Index than Kimi Code CLI
DeepSWE
69%
64%
Codex has a higher DeepSWE score than Kimi Code CLI
Terminal-Bench v2
88%
84%
Codex has a higher Terminal-Bench v2 score than Kimi Code CLI
SWE-Atlas-QnA
43%
37%
Codex has a higher SWE-Atlas-QnA score than Kimi Code CLI
Cost per Task
$7.08
$3.18
Kimi Code CLI has a lower cost per task than Codex
Time per Task
10.2m
23.8m
Codex has a lower time per task than Kimi Code CLI
Turns per Task
114.2
124.6
Codex has a lower turns per task than Kimi Code CLI
Token Usage per Task
13.2M
10.6M
Kimi Code CLI has a lower token usage per task than Codex
Cache Hit Rate
90%
95%
Kimi Code CLI has a higher cache hit rate than Codex

Model Variants

Evaluated model variants for Codex and Kimi Code CLI.

Model Variants

Codex
GPT-5.6 Sol (max)
67
69%
88%
43%
$7.08
10.2m
13.2M
Codex
GPT-5.6 Sol (xhigh)
65
67%
86%
42%
$5.24
7.4m
9.9M
Codex
GPT-5.6 Sol (high)
64
65%
83%
45%
$4.14
6.3m
8.1M
Codex
GPT-5.6 Terra (max)
62
67%
84%
36%
$2.21
8.4m
9.5M
Codex
GPT-5.5 (xhigh)
61
64%
84%
36%
$5.07
10.1m
12.3M
Codex
GPT-5.6 Sol (medium)
61
64%
78%
40%
$2.99
5.2m
5.8M
Codex
GPT-5.6 Luna (max)
59
63%
80%
33%
$0.31
8.0m
15.5M
Codex
GPT-5.6 Terra (xhigh)
57
58%
81%
32%
$1.52
6.9m
6.5M
Codex
GPT-5.6 Terra (high)
56
60%
76%
31%
$1.27
6.2m
5.5M
Codex
GPT-5.6 Luna (xhigh)
55
57%
76%
31%
$0.25
6.6m
12.3M
Codex
GPT-5.5 (medium)
54
57%
76%
31%
$2.75
6.4m
7M
Codex
GPT-5.6 Sol (low)
54
53%
73%
34%
$1.72
3.7m
3.2M
Codex
GPT-5.6 Luna (high)
51
53%
72%
29%
$0.19
5.7m
9.5M
Codex
GPT-5.6 Terra (medium)
48
46%
69%
28%
$0.72
4.3m
3.1M
Codex
GPT-5.6 Sol (none)
43
35%
61%
34%
$1.40
3.4m
3.4M
Codex
GPT-5.6 Luna (medium)
42
37%
63%
27%
$0.09
3.4m
4.4M
Codex
GPT-5.4 (medium)
39
25%
70%
22%
$2.42
7.1m
5.9M
Codex
GPT-5.6 Terra (low)
37
30%
58%
23%
$0.39
2.8m
1.5M
Codex
GPT-5.6 Luna (low)
25
10%
50%
15%
$0.04
1.9m
1.5M
Codex
GPT-5.6 Terra (none)
24
13%
39%
19%
$0.30
1.8m
1.1M
Codex
GPT-5.6 Luna (none)
20
6%
37%
17%
$0.07
2.5m
3.6M
Kimi Code CLI
Kimi K3
61
64%
84%
37%
$3.18
23.8m
10.6M

Performance

Performance across the Artificial Analysis Coding Agent Index.

Artificial Analysis Coding Agent Index

Artificial Analysis Coding Agent Index v1.3 incorporates 3 benchmarks: DeepSWE, Terminal-Bench v2, and SWE-Atlas-QnA · Higher is better
Since benchmarking, we have observed a higher rate of content safety filtering on this endpoint.

The Artificial Analysis Coding Agent Index is a composite score built from DeepSWE, Terminal-Bench v2, and SWE-Atlas-QnA.

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.

Token Usage

Token consumption across the Artificial Analysis Coding Agent Index.

Token Usage per Task

Average input, cache, and output tokens per task
Prompt cache hit rates can vary significantly by provider routing, which can materially change effective cost.

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. Total Tokens

Artificial Analysis Coding Agent Index vs. average total tokens per task
Most attractive quadrant

Each point represents a coding-agent variant. Farther right means higher benchmark performance, while lower token usage appears farther left. Agents toward the upper-left use fewer tokens for a given level of performance.

Cost

Pay-per-token API cost across the Artificial Analysis Coding Agent Index, based on current per-token pricing.

Cost per Task

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

This chart shows the average pay-per-token API cost per task across the Artificial Analysis Coding Agent Index, spanning DeepSWE, Terminal-Bench v2, and SWE-Atlas-QnA.

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.

Artificial Analysis Coding Agent Index vs. Cost per Task

Artificial Analysis Coding Agent Index vs. average pay-per-token API cost per task (USD)
Most attractive quadrant
Pareto line

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

Execution Time

Active agent runtime across the Artificial Analysis Coding Agent Index.

Time per Task

Average agent wall time per task · Lower is better

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. Execution Time

Artificial Analysis Coding Agent Index vs. average agent wall time per task
Most attractive quadrant

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