Clay AI agent benchmark: how well Kimi K3 works with Clay

With Clay, Kimi K3 posted the two highest job change averages of seven agents in our benchmark: 94.7 of 100 in Factory and 93.7 in OpenCode, at about 14 Clay credits per person in five of their six runs. In both harnesses it found 93 of 100 work emails in every run, once for 334.5 credits, the lowest of any work email run. Across all five GTM data tasks, Factory with Kimi K3 averaged 83.0, within 0.2 points of the top agent.

By The Prompting Company · Snapshot November 5, 2026

What does Kimi K3 do well with Clay?

We ran Kimi K3 in two harnesses, OpenCode and Factory, through five GTM data tasks with Clay, three runs each. We call them OpenCode Kimi and Factory Kimi below. Clay credits are measured as the change in the Clay workspace credit balance.

Job changes. Both harnesses ran Clay's Enrich Person and Find Contact Details on all 100 LinkedIn profiles in every run, caught 42 of the 45 people who had moved, and scored 93 to 96. Factory Kimi's second run scored 96, the best single job change run in the benchmark. OpenCode Kimi spent 1,404 to 1,406 credits per run, about 14 per person, and finished in 8.6 to 13.5 minutes.

Work emails. Both found 93 of 100 verified emails in all six runs, the top score on the task. Factory Kimi's first run spent 334.5 credits, the lowest of all 21 work email runs, and OpenCode Kimi had the fastest median of any agent, 5.1 minutes. With Clay's Work Email waterfall, you only pay credits for the provider that finds a match.

Company data. Factory Kimi's first run was the only run in the benchmark to score 100 on domain, HQ country and headcount, and its 79 on funding rounds was the best single funding score. It confirmed companies with Clay's people and company search and checked each news article's date as well as its headline. OpenCode Kimi's first run confirmed matches in Clay's company search and scored 98 on domain, 99 on HQ and 99 on headcount for 324 credits, about 3.2 per company.

CMOs. Factory Kimi passed all three of its CMO runs. In the first, it used Clay's people search to choose each CMO, then ran Work Email for that one person: 80 of 100 for 312 credits, about 3.1 per company.

Target lists. Factory Kimi's first run listed 74 qualifying companies and passed. OpenCode Kimi built a list with 66 qualifying companies from Clay's company search for 7 credits.

The best setup for Kimi K3 with Clay

Pick the harness for the job. Factory Kimi fits multi-step research such as CMOs and company data, where it passed every CMO run and set the top company data run. OpenCode Kimi fits quick per-record work: its work email runs took 3.7 to 8.7 minutes, its job change runs 8.6 to 13.5, and its model cost averaged $1.07 per reported run against Factory Kimi's $1.29.

Name the Clay method in the prompt. In either harness:

1. Work emails: run Work Email once per contact.
2. Job changes: run Enrich Person and Find Contact Details once per LinkedIn URL and decide every row.
3. CMOs: choose the person with Find People, which uses 0 Data Credits, then run one Work Email lookup.
4. Company data: confirm each company in Clay's company search, and check the date on each funding article.
5. Target lists: confirm each company is listed in its own name and headquartered in the US before adding it.

Write the output once every row has a Clay answer. Running extra lookups added credits without improving results.

Kimi K3 results across the five tasks

Primary score out of 100, averaged over three runs.

*Includes one run that reached the 60-minute limit and counts as 0.

Factory Kimi passed 11 of its 15 runs, the second most of seven agents. For one average run of all five tasks, OpenCode Kimi spent 3,560 Clay credits and Factory Kimi 3,541. For more on how the two harnesses compare, see "Does the agent harness matter when using Clay with the same model?".

How we tested

Seven agents ran each of five tasks three times, one agent at a time, in a fresh sandbox with the same Clay key, the same 19 saved Clay routines and a 60-minute limit. A timeout counts as 0, and platform failures were rerun. Each task had 100 records. Model cost is as reported by the test platform. Tasks, prompts and graders are at github.com/promptingcompany/clay-bench.

Try it yourself

The Clay Agent Plugin ships Clay's skills to your agent up front, including the habits that keep credit use down; setup is one prompt (about 15 minutes), and there is no additional cost to use it. Install the Clay Agent Plugin in OpenCode or Factory, or give Kimi K3 a Clay API key and the Clay Public API docs. Add the five method lines above to your prompt. Agent work uses the same credits as in-product work (Clay pricing).

Common questions

How well does Kimi K3 work with Clay? Very well on job changes and work emails: Kimi K3 posted the two top job change averages, 94.7 in Factory and 93.7 in OpenCode, and found 93 of 100 work emails in every run.

Which harness should I use for Kimi K3 with Clay? Factory for multi-step research such as CMOs and company data, and it passed 11 of 15 runs overall; OpenCode for fast per-record tasks at a lower model cost per run.

How many Clay credits does Kimi K3 use? About 14 credits per person for a job change check, as little as 334.5 credits for 100 work email contacts, and about 3.1 per company in its best CMO run.

How does Kimi K3 compare with closed models on Clay? Factory Kimi's 83.0 average was within 0.2 points of the top closed-model setup. See "Open-weight or closed models: which works better with Clay?".