Moonshot · tested in Taskade Genesis
Kimi K3 in Taskade
Kimi K3 from Moonshot runs on the Max and Enterprise plans. We last tested it on Sep 30, 2026 in TSK-1, our app-building test.
- Status
- Available
- Plan
- Max and Enterprise
- Where to pick it
- More models or All Models
- Your own key
- Together AI
- Last test
- Context window
- 1.05Mtokens
- Max output
- 1.05Mtokens
- Released
- Where it runs
- 15services
TSK-1 scorecard for the Kimi family
TSK Score
Kimi family, across real apps made in Taskade Genesis
- Interfaceready to share
- Strong
- Taskcompletes your task
- Strong
- Memorykeeps your data
- Strong
- Adapthandles follow-up edits
- Strong
Builds more than almost any model and finishes later than all of them.
MoreLess
Its sales CRM scored all eight sample leads and its match tracker was the richest of its test.
Published scores
Benchmark scores that Moonshot and test sites published for Kimi K3. Taskade did not re-run them. Lab-reported marks a score from Moonshot itself.
| Benchmark | Score | Source |
|---|---|---|
| Terminal-Benchaccuracy · v2.1 · max effort · Kimi Code · KimiLab-reported | 88.3% | KimiLab-reported |
| BrowseCompaccuracy · max effort, context compaction · KimiLab-reported | 91.2% | KimiLab-reported |
More published scores (11)
| Benchmark | Score | Source |
|---|---|---|
| DeepSWEresolve rate · v1.1 · max effort · Kimi Code · KimiLab-reported | 67.5% | KimiLab-reported |
| FrontierSWEdominance score · max effort · Kimi Code · KimiLab-reported | 81.2 | KimiLab-reported |
| Program Benchmax effort · Kimi Code · KimiLab-reported | 77.8 | KimiLab-reported |
| SWE Marathonresolve rate · v1.1 · max effort · Claude Code · KimiLab-reported | 42% | KimiLab-reported |
| GDPval-AAElo · v2 · max effort · KimiLab-reported | 1,668 | KimiLab-reported |
| AA-BriefcaseElo · max effort · KimiLab-reported | 1,548 | KimiLab-reported |
| AutomationBenchsuccess rate · max effort · 600-task public subset · KimiLab-reported | 30.8% | KimiLab-reported |
| JobBenchmax effort · KimiLab-reported | 52.9 | KimiLab-reported |
| SpreadsheetBenchv2 · max effort · Claude Code · KimiLab-reported | 34.8 | KimiLab-reported |
| CharXiv Reasoningaccuracy · max effort, with tools · KimiLab-reported | 91.3% | KimiLab-reported |
| ZeroBenchpass@5 · max effort, with tools · KimiLab-reported | 41% | KimiLab-reported |
LMArena scores
| Benchmark | Score | Source |
|---|---|---|
| Text arena±5 · 32,795 votes · Rank 18 of 416 · LMArena | 1,487 | LMArena |
| WebDev arena±6 · 19,426 votes · Rank 15 of 145 · LMArena | 1,655 | LMArena |
Scores from the LMArena leaderboard dataset (CC BY 4.0). People vote between two answers, and the votes set the score. Taskade picked the Overall category and the default setting of each model.
Scores collected by models.dev (MIT). Each source links the page that published the score.
Versions tested
Open a version to read its findings.
Kimi K3Tested Sep 20266 findings
Builds more than almost any model and finishes later than all of them. Its sales CRM scored all eight sample leads and its match tracker was the richest of its test.
The most reliable model of the test: only 7.2% of its build actions went wrong, and it needed the fewest actions. Its closing summary accurately described the app it had built.
A good-looking, complete match tracker with a heroes page; we could not confirm a saved match from outside the app.
Built a complete client sign-up app with 31 of 32 questions word for word, then reached the 40-minute limit before a closing summary.
Its sales CRM scoring automation ran eight times and scored all 8 sample leads.
Built the richest match tracker of the test, with four pages, heroes, and recaps.
On a real customer's family intake form it wrote no app page within the 15-minute test limit.
Model facts
The token limits are the most the model accepts. Taskade can run it with lower limits.
- Last updated
- Input types
- Text, Images, Video
- Output types
- Text
- Reasoning
- Yes
- Tool calling
- Yes
- File attachments
- Yes
- Structured output
- Yes
- Open weights
- Yes
- Reasoning effort
- LowHighMax
Published specs from the model maker, via models.dev (MIT).
Run Kimi K3 with your own key
Taskade runs it for you. You can also pick it with your own key in a custom agent.
| Key provider | Plan |
|---|---|
| Together AI | Business and up |
Where it runs
Services outside Taskade that serve Kimi K3, from models.dev. Each link opens in a new tab. In Taskade, your own key works only through the key providers on this page.
How to choose Kimi K3
- In the builder chat, open the model menu and choose More models. Then pick the model.
- In the agent editor, pick the model under All Models.
- If your model menu does not show the model, Taskade uses TSK-Auto instead.
Where Kimi ranks
TSK Score out of 100.
Compare Kimi
Other Moonshot models
FAQ
Which Taskade plans include Kimi K3?
Kimi K3 runs in Taskade on the Max and Enterprise plans. If your model menu does not show it, Taskade uses TSK-Auto instead.
Has Taskade tested Kimi K3?
Yes. We last tested Kimi K3 on Sep 30, 2026 in the TSK-1 Benchmark. Builds more than almost any model and finishes later than all of them. Its sales CRM scored all eight sample leads and its match tracker was the richest of its test.
How do I choose Kimi K3 in Taskade?
In the builder chat, open the model menu and choose More models. Then pick Kimi K3. In the agent editor, pick it under All Models.
Can I use my own API key for Kimi K3?
Yes. In a custom agent, you can pick Kimi K3 with your own key from Together AI (Business and up). Your provider bills the model use to your own account.
Who makes Kimi K3?
Moonshot makes Kimi K3. Taskade runs it on the Max and Enterprise plans, beside frontier models from other top AI labs.