Kimi K3
Moonshot AI's open-weight model for long coding, research and document work.
Made on Runable
Each clip is one prompt, run as a single Runable task.
Kimi K3 on Runable
Long sessions
Made for long coding and research jobs with little hand-holding.
Reads your files
Hand it PDFs, spreadsheets and screenshots.
No Moonshot account
Runs on your Runable credits, next to other models.
How to use Kimi K3 on Runable
Select Kimi K3
Start a task on Runable and pick Kimi K3 in the model menu.
Describe the job
Type what you need, or add files and links. Runable plans it and gets to work.
Get the finished work
Refine by chatting, then publish the website, download the video or export the deck.
One prompt, finished work
Real results from single Runable tasks. Pick one to watch it.
- token context window
- 1M
- parameters, open weights
- 2.8T
- SWE Marathon, Moonshot-reported
- 42.0
- turn to build
- Your
Why choose Kimi K3
Open frontier weights
2.8T parameters, with weights on Hugging Face.
Long engineering runs
42.0 on SWE Marathon, top of Moonshot’s launch table.
Business workflows
30.8 on AutomationBench, also top of that table.
Huge context
1M tokens for codebases, reports and long PDFs.
Sees what it builds
Native vision lets it check screenshots as it works.
Lower cost per task
$2.00 per Artificial Analysis task, against $5.46 for Sonnet 5.5.
Kimi K3 specs
- Maker
- Moonshot AI
- Model
- Kimi K3 (followed by Kimi K3.1 Pro in October 2026)
- Released
- July 16, 2026
- API model ID
- kimi-k3
- Context window
- 1M tokens
- Max output
- 131K tokens by default (up to 1M on Kimi’s API)
- Input / output
- Text and images in, text out
- Reasoning
- Always on; effort low, high or max (default max)
- Architecture
- Mixture-of-experts, 2.8T total / 104B active parameters
- API price
- $3 input / $15 output per 1M tokens ($0.30 cached input)
Kimi K3 vs K3.1 Pro and similar picks
Kimi K3.1 Pro is Moonshot’s newer, post-trained version of K3. Speeds are from Artificial Analysis, October 2026; prices are each maker’s API list prices.
| Model | Best for | Speed | API price in / out |
|---|---|---|---|
| Kimi K3 | Long coding and research sessions | 41 tokens/s | $3 / $15 |
| Kimi K3.1 Pro | Moonshot’s newer flagship | 74 tokens/s | $2 / $10 |
| Sonnet 5.5 | Everyday tasks, best speed for the smarts | 134 tokens/s | $2 / $10 |
| Grok 4.6 | Low-cost first drafts and agent work | 80 tokens/s | $2 / $6 |
The only AI agent you need
Need SSO, data localization, premium support, onboarding assistance? Contact Sales
Put Kimi K3 to work
Describe the job. Runable runs it on Kimi K3 and ships the result.






