AI Models
Kimi K3, Online in MoClaw
Run Kimi K3 online in MoClaw's managed AI computer — no self-hosting, no API keys. Pick Kimi K3, type a task, and go.
MoClaw is an independent managed AI computer where you can run frontier models with no setup or self-hosting.
All integrationsWhat Kimi K3 Is Built For
The kind of work Kimi K3 is made for.
Build working web apps and games
Kimi K3 turns ideas into interactive web apps and 3D games — the build-and-preview loop MoClaw runs end to end.
Try itDeep research into dashboards
It reviews sources at scale and turns findings into interactive dashboards and charts, not just a wall of text.
Try itLong-horizon, multi-step projects
Built for agentic autonomy: planning across stages, delegating to subagents, and checking its own work over long runs.
Try itData analysis and visualization
Point it at a dataset and it profiles, analyzes, and generates the visualizations that explain what matters.
Try itSee Kimi K3 in action
Live demos built with Kimi K3, embedded from the source — the maker's demos, not MoClaw output.
A procedural 3D open-world game
Kimi K3 generated this browser-based 3D open world — forests, a log-cabin village, snowy mountains, and dynamic weather — with Three.js and WebGPU compute and no game engine, a showcase of the model's long-horizon coding.
A working macOS desktop, in the browser
A macOS-style desktop rebuilt to run in the browser by Kimi K3, with a working dock, widgets, and windows — an example of the front-end and agentic UI work the model handles end to end.
How it works
3 steps to wire up Kimi K3, no engineering required.
- 1
Kimi K3 is on MoClaw
Kimi K3 is available in MoClaw's managed AI computer. Pick it in any chat, right next to the models you already use — no waitlist, no setup, no separate key.
- 2
Or let Auto route to it
On Auto, MoClaw picks the right model per turn and routes to Kimi K3 when a task plays to its strengths, with no configuration from you.
- 3
Switch mid-conversation
Switch to Kimi K3 in the middle of a thread and the rest keeps working — your files, history, and prompt caching carry over, with no re-setup.
Why Run Kimi K3 Inside MoClaw
What is Kimi K3?
Kimi K3 is Moonshot AI's newest model, announced on July 16, 2026, with open weights. It is a 2.8-trillion-parameter, natively multimodal model with a 1-million-token context window, which Moonshot describes as the first open 3T-class model. Moonshot targets Kimi K3 at long-horizon coding, knowledge work, and reasoning, and reports benchmark scores including 88.3 on Terminal Bench 2.1 and 93.5 on GPQA-Diamond.
Moonshot's own demos show Kimi K3 building interactive web apps and 3D games, running multi-step research pipelines with parallel subagents, and turning findings into interactive dashboards and motion-graphics explainers. Those are Moonshot's showcases, not MoClaw output — but they are exactly the long-horizon, agentic tasks MoClaw is built to run.
Running Kimi K3 on MoClaw
MoClaw is a managed AI computer: a persistent cloud machine with a browser, a file system, and tools already wired up. Point a task at Kimi K3 and MoClaw runs the work end to end, instead of you standing up an inference server, managing keys, and gluing tools together. Pick Kimi K3 in any chat, or let Auto route to it — your files, history, and credits carry across every model.
Kimi K3 pricing and access
On Moonshot's own platform, Kimi K3 is listed at $0.30 per million cached-input tokens, $3.00 per million input tokens, and $15.00 per million output tokens. On MoClaw, Kimi K3 runs on your existing plan credits with no separate provider bill to manage and no key to wrangle; BYOK remains an option for models that support it.
Kimi K3 on MoClaw vs running Kimi K3 yourself
Pick the one that fits how you use AI.
| MoClaw (managed) | Self-hosted / DIY | |
|---|---|---|
| Setup | None — Kimi K3 is a menu choice in any chat. | Provision GPUs or wire the Kimi API and manage keys. |
| Tools | Browser, files, code, and schedules already connected. | Build and maintain your own tool layer. |
| Persistence | Files and history stay in your cloud computer. | You manage state and storage yourself. |
| Switching models | Swap Kimi K3 for another frontier model in one click. | Re-plumb per provider. |
Pairs with
Integrations people commonly connect alongside Kimi K3.
FAQ
Quick answers about pricing, privacy, and limits.
- Is Kimi K3 available on MoClaw?
- Yes. Kimi K3 runs inside MoClaw's managed AI computer — pick it in any chat, with no self-hosting and no API key required. It is covered by your MoClaw plan.
- What is Kimi K3?
- Kimi K3 is the newest model from Moonshot AI, announced July 16, 2026: a 2.8-trillion-parameter, natively multimodal model with a 1-million-token context window, positioned as the first open 3T-class model and built for long-horizon coding, knowledge work, and reasoning.
- How much does Kimi K3 cost?
- On Moonshot's platform Kimi K3 is listed at $0.30 per million cached-input tokens, $3.00 per million input tokens, and $15.00 per million output tokens. On MoClaw, Kimi K3 runs on your monthly plan credits with no markup, so you can use it without metering an API account yourself.
- Can I use Kimi K3 without self-hosting?
- Yes. Kimi K3 runs inside MoClaw's managed AI computer, so there is nothing to self-host and no key to wrangle. If you would rather meter usage on your own account, BYOK remains an option for models that support it.
- What can I build with Kimi K3 on MoClaw?
- Kimi K3 is built for long-horizon coding and agentic work — building web apps and games, recreating sites from a screenshot, running multi-step research into dashboards, and analyzing data. On MoClaw it runs those end to end, with a browser, files, and tools already connected.
- Is Kimi K3 open source?
- Yes — Moonshot released Kimi K3 as an open-weights model, which it calls the first open 3T-class model. On MoClaw you can use Kimi K3 as a managed, hosted model without downloading or self-hosting those weights yourself.
- What is Kimi K3's context window?
- Kimi K3 ships with a 1-million-token context window, so it can work over large repositories, long documents, and multi-file projects in a single session.
- Is Kimi K3 good for coding and agentic tasks?
- That is its focus. Moonshot demos Kimi K3 on long-horizon coding and agentic work — writing and optimizing GPU kernels, building compilers from scratch, and running multi-step projects with subagents. On MoClaw those are exactly the workflows the managed AI computer is built to run.
- Does Kimi K3 support images and vision?
- Yes. Kimi K3 is natively multimodal, so it can reason over screenshots and images in the loop, not just text.
Related reading
Guides that go deeper on Kimi K3 and AI automation.
What Is Kimi K3? Moonshot's 2.8T Model
What is Kimi K3? Moonshot's 2.8T-parameter flagship launched July 16, 2026 with a 1M context window and open weights due July 27. Specs, price, access.
Kimi K3 vs Claude: Benchmarks & 2026 Verdict
Kimi K3 vs Claude: K3 claims wins over Opus 4.8, but Fable 5 still leads overall. Benchmarks, per-task costs, and which model to pick in 2026.
Kimi K3 Agent Swarm: 300 Parallel Agents
Kimi K3 Agent Swarm coordinates up to 300 sub-agents and 4,000 tool calls per task. How the architecture works, where it wins, and where it breaks.
Kimi K3 vs K2.6: What Changed & Upgrade?
Kimi K3 vs K2.6: 2.8x the parameters, 4x the context, and a 5x price jump. What K3 adds, what K2.6 still does better, and who should upgrade in 2026.