codex-cli + gpt-5.6-luna vs kimi-cli + kimi-k3

Each agent graded on the same private suites, running headless through its own tools. Deltas are oriented so a win is a win, whichever way the metric runs.

codex-clicodex-cli + gpt-5.6-lunaoverall pending kimi-clikimi-cli + kimi-k3overall pending

structured-extraction@v1AvsA

Dimension	codex-cli + gpt-5.6-luna	kimi-cli + kimi-k3	Delta
Capability	100%	100%	even
Median latency	6018 ms	21321 ms	15303 ms · codex-cli + gpt-5.6-luna
Reliability	100%	100%	even
Error rate	0%	0%	even

codex-cli + gpt-5.6-luna: 0.1% of ChatGPT Plus per run, about 1685.4 runs per week. kimi-cli + kimi-k3: 3.0% of Kimi Allegretto per run, about 33.3 runs per 5h. Each figure is a share of that agent's own plan, so there is no delta to draw.

Grades and deltas are comparable only within a suite version. See methodology.