Moonshot AI offers Kimi, a 1‑trillion‑parameter mixture‑of‑experts language model that supports up to 256 k token context and multimodal inputs (text, image, audio). The platform provides real‑time web search, code execution, and file‑based Q&A through a unified API, enabling developers to embed extended‑context, reasoning, and tool‑oriented capabilities with usage‑based billing. Services are cloud‑hosted with enterprise‑grade security and monitoring.
Funding
$4.6B raised to dateRaised to date based on public sources. This may differ from the amount the company actually raised and is based only on what is publicly available on the internet.
AGAH+7Founders
Product
Problem
Users and developers often encounter AI assistants that are limited to short context windows, lack reliable web search integration, and provide minimal multimodal capabilities, making it difficult to handle complex queries, large documents, or tasks that require image, audio, or tool assistance.
Solution
Moonshot AI delivers Kimi, a large language model‑driven AI assistant that supports real‑time web search, deep reasoning, and multimodal inference across text, images, and audio. The model operates with an extended context window of up to 256 k tokens, enabling coherent long‑form interactions. Kimi’s architecture uses a 1‑trillion‑parameter Mixture‑of‑Experts (MoE) backbone with 32 B activation parameters, delivering benchmark‑level performance on reasoning, coding, and agent tasks. An open platform provides flexible API endpoints for chat, tool use, partial responses, file‑based Q&A, and streaming output, allowing developers to embed advanced AI capabilities with minimal integration effort. All services are hosted in the cloud with usage‑based billing and secure data handling.
Target Audience
The primary customers are consumer users seeking a high‑performance AI assistant and software developers or enterprises that need to integrate advanced LLM capabilities—such as extended context, multimodal input, and tool use—into their applications, platforms, or internal workflows.
Features
- MoE‑based 1 T parameter model (Kimi K2) with 32 B activation parameters and 256 k token context length
- Integrated web‑search tool, code execution, and file‑based question answering via unified API calls
- Multimodal support for image, audio, and text inputs, enabling vision and speech understanding
- Agent‑oriented capabilities for multi‑step tool orchestration and autonomous reasoning
- Comprehensive developer suite: RESTful chat, tool‑use, partial mode, JSON mode, streaming, CLI, and web playground
- Fine‑grained usage monitoring, API key management, and token‑estimate utilities in the console
- Enterprise‑grade security with encrypted data transmission and role‑based access controls