Transparent pricing with no hidden fees. Pay as you go.
kimi-k3kimi-k3| Model | Spec | PoYo price | Official price | You save |
|---|---|---|---|---|
| kimi-k3 | Input tokens | $2.28/1M tokens 456 credits /1M tokens | $3.00/1M tokens Official | 24% |
Output tokens | $11.40/1M tokens 2280 credits /1M tokens | $15.00/1M tokens Official | 24% |
* Actual fees are based on the final output.
Complete guide to using Kimi K3 API: Open Frontier Intelligence for Coding and Knowledge Work
Kimi K3 is Moonshot AI's 2.8-trillion-parameter flagship model, built for long-horizon coding, deep reasoning, and end-to-end knowledge work. Its native multimodal understanding and 1M-token context let agents work across large codebases, documents, images, video, and complex tool-driven tasks.
Kimi K3 is available on PoYo through the kimi-k3 model ID and the OpenAI-compatible Chat Completions API.
Kimi K3 combines extreme model scale, native multimodal understanding, long context, and agentic execution for demanding coding and knowledge workflows.
01
Kimi K3 is designed to sustain extended engineering sessions with limited supervision. It can navigate large codebases, coordinate terminal tools, and use visual feedback to refine software.
02
Text, screenshots, images, and video can be understood within one model, enabling vision-in-the-loop agents that inspect results and improve their next action.
03
A one-million-token context window gives Kimi K3 room for large evidence sets, long interaction histories, source code, and multi-stage deliverables without prematurely discarding context.
04
Kimi K3 supports custom tools, required tool choice, dynamic tool loading, streaming, and strict JSON Schema responses for application workflows.
Run extended coding tasks across large repositories, coordinate terminal tools, debug systems, and carry implementation work from investigation to verification.
Combine code with screenshots and visual feedback to iterate on frontends, playable games, 3D scenes, CAD workflows, and interactive experiences.
Synthesize large document collections, connect evidence, write research code, and produce structured reports or interactive visualizations.
Build agent workflows for consulting reports, financial analysis, slides, spreadsheets, dashboards, and multi-step business deliverables.
Create agents that call custom functions, load tools dynamically, return structured JSON, and keep working across long multi-turn sessions.
Analyze visual inputs alongside text for UI review, media understanding, technical diagrams, visual QA, and multimodal reasoning.
These specifications come from Moonshot AI's official Kimi K3 launch materials. PoYo pricing is 456 input credits and 2,280 output credits per million tokens.
| Capability | Kimi K3 | What It Means |
|---|---|---|
| Model scale | 2.8 trillion parameters | An open 3T-class Mixture-of-Experts model |
| Context window | 1 million tokens | Work across large repositories and document sets |
| Modalities | Native text, image, and video understanding | Reason over code, documents, screenshots, and media |
| Reasoning | Always-on thinking with low, high, and max effort | Match reasoning depth to task complexity |
| Agent features | Tool calling and dynamic tool loading | Build long-running, tool-driven workflows |
| Structured output | Strict JSON Schema support | Generate reliable application-ready responses |
| Architecture | KDA, AttnRes, Stable LatentMoE | Efficient information flow at extreme scale |
| PoYo availability | Available | Use model ID kimi-k3 with Chat Completions |
PoYo provides Kimi K3 through an OpenAI-compatible API. Current pricing is shown in the pricing section above.
Call Kimi K3 now with a confirmed model ID, transparent token pricing, and OpenAI-compatible requests.
Compare Kimi K3 with other coding, reasoning, and multimodal models without rebuilding your discovery workflow.
Use the interactive playground and practical API documentation to move from evaluation to production.
The model ID, endpoint, availability, and token prices are aligned with PoYo's production configuration.
