# Moonshot AI

> The lab behind Kimi K3, the most capable open-weight model, with 1M context.

Canonical: https://www.subconscious.dev/providers/moonshot-ai · By The Subconscious Team · Updated September 30, 2026

- Founded: 2023
- Example models: Kimi K3, Kimi K2.6
- Website: https://www.moonshot.ai

## Overview

Moonshot AI is the Beijing lab behind the Kimi models. Its flagship Kimi K3 launched July 16, 2026 as a 2.8 trillion parameter mixture-of-experts model that activates 16 of 896 experts per token, with native vision and a 1M token context. It is the first open model in the 3T class, and full weights landed on Hugging Face on July 27. The hosted API costs $3 in and $15 out per million tokens, with cached input at $0.30, and it runs through an OpenAI-compatible endpoint, Kimi Code in the terminal, OpenRouter and Cloudflare Workers AI.

Independent testers mostly confirmed Moonshot's claims. Vals AI scored K3 at 93.4% on SWE-bench Verified with a neutral harness, fourth overall behind Claude Opus 5, GPT-5.6 Sol and Claude Fable 5. Artificial Analysis placed it third on its Intelligence Index. The catch is speed and verbosity: it always thinks, and it runs around 33 tokens per second. Demand overran Moonshot's GPUs within days of launch, and new API subscriptions paused on July 19 before reopening in batches. The cheaper Kimi K2.6 remains available at $0.95 in and $4 out.

## Upsides

- Most capable open-weight model, near closed frontier coding scores at a lower price.
- Very large context and strong cache discounts for repo-scale agents.

## Core use cases

- Long-horizon coding on huge repositories.
- Document-heavy research and visual agent work that needs 1M context.

## Downsides

- Slow and verbose, and self-hosting takes a 64+ accelerator cluster.
- The custom license adds a commercial agreement above $20M in hosting revenue and a branding clause at large scale.

## At a glance

| Attribute | Value |
|---|---|
| Model access | Open weights, custom license |
| Flagship models | Kimi K3, Kimi K2.6 |
| Speed | ~33 tok/s on Kimi K3 |
| Price | $3 in, $15 out (Kimi K3) |
| Customization | Open weights to fine-tune |
| Deployment | API, Kimi Code, OpenRouter |
| Long context | 1M |

## FAQ

### What is Moonshot AI?

Moonshot AI is the Beijing lab behind the Kimi models. Its flagship Kimi K3 launched July 16, 2026 as a 2.8 trillion parameter mixture-of-experts model that activates 16 of 896 experts per token, with native vision and a 1M token context. It is the first open model in the 3T class, and full weights landed on Hugging Face on July 27. The hosted API costs $3 in and $15 out per million tokens, with cached input at $0.30, and it runs through an OpenAI-compatible endpoint, Kimi Code in the terminal, OpenRouter and Cloudflare Workers AI.

### What is Moonshot AI best for?

Long-horizon coding on huge repositories; Document-heavy research and visual agent work that needs 1M context.

### How much does Moonshot AI cost?

Moonshot AI pricing at a glance: $3 in, $15 out (Kimi K3). Rates change often, so check Moonshot AI's pricing page before committing.

### How much context does Moonshot AI support?

Moonshot AI's long-context support: 1M.

### What are the downsides of Moonshot AI?

Slow and verbose, and self-hosting takes a 64+ accelerator cluster; The custom license adds a commercial agreement above $20M in hosting revenue and a branding clause at large scale.

### What are the best alternatives to Moonshot AI?

Common alternatives include Subconscious, OpenAI, Anthropic, Google Vertex AI, Amazon Bedrock. Each has a head-to-head comparison with Moonshot AI on this site.

## Comparisons

- [Subconscious vs Moonshot AI](https://www.subconscious.dev/compare/subconscious-vs-moonshot-ai.md)
- [OpenAI vs Moonshot AI](https://www.subconscious.dev/compare/openai-vs-moonshot-ai.md)
- [Anthropic vs Moonshot AI](https://www.subconscious.dev/compare/anthropic-vs-moonshot-ai.md)
- [Google Vertex AI vs Moonshot AI](https://www.subconscious.dev/compare/google-vertex-vs-moonshot-ai.md)
- [Amazon Bedrock vs Moonshot AI](https://www.subconscious.dev/compare/aws-bedrock-vs-moonshot-ai.md)
- [Together AI vs Moonshot AI](https://www.subconscious.dev/compare/together-ai-vs-moonshot-ai.md)
- [Fireworks AI vs Moonshot AI](https://www.subconscious.dev/compare/fireworks-vs-moonshot-ai.md)
- [Baseten vs Moonshot AI](https://www.subconscious.dev/compare/baseten-vs-moonshot-ai.md)
- [Groq vs Moonshot AI](https://www.subconscious.dev/compare/groq-vs-moonshot-ai.md)
- [Cerebras vs Moonshot AI](https://www.subconscious.dev/compare/cerebras-vs-moonshot-ai.md)
- [DeepInfra vs Moonshot AI](https://www.subconscious.dev/compare/deepinfra-vs-moonshot-ai.md)
- [Modal vs Moonshot AI](https://www.subconscious.dev/compare/modal-vs-moonshot-ai.md)
- [xAI vs Moonshot AI](https://www.subconscious.dev/compare/xai-vs-moonshot-ai.md)
- [DeepSeek vs Moonshot AI](https://www.subconscious.dev/compare/deepseek-vs-moonshot-ai.md)
- [Moonshot AI vs Z.ai](https://www.subconscious.dev/compare/moonshot-ai-vs-z-ai.md)
- [Moonshot AI vs Alibaba Cloud](https://www.subconscious.dev/compare/moonshot-ai-vs-alibaba-cloud.md)
- [Moonshot AI vs Meta](https://www.subconscious.dev/compare/moonshot-ai-vs-meta.md)
- [Moonshot AI vs SambaNova](https://www.subconscious.dev/compare/moonshot-ai-vs-sambanova.md)
- [Moonshot AI vs Nebius](https://www.subconscious.dev/compare/moonshot-ai-vs-nebius.md)
- [Moonshot AI vs fal](https://www.subconscious.dev/compare/moonshot-ai-vs-fal.md)
- [Moonshot AI vs Novita AI](https://www.subconscious.dev/compare/moonshot-ai-vs-novita-ai.md)
- [Moonshot AI vs Parasail](https://www.subconscious.dev/compare/moonshot-ai-vs-parasail.md)
- [Moonshot AI vs Inference.net](https://www.subconscious.dev/compare/moonshot-ai-vs-inference-net.md)
- [Moonshot AI vs GMI Cloud](https://www.subconscious.dev/compare/moonshot-ai-vs-gmi-cloud.md)
- [Moonshot AI vs Sail Research](https://www.subconscious.dev/compare/moonshot-ai-vs-sail-research.md)
- [Moonshot AI vs Morph](https://www.subconscious.dev/compare/moonshot-ai-vs-morph.md)
- [Moonshot AI vs Relace](https://www.subconscious.dev/compare/moonshot-ai-vs-relace.md)
- [Moonshot AI vs TypeSafe AI](https://www.subconscious.dev/compare/moonshot-ai-vs-typesafe-ai.md)
- [Moonshot AI vs StepFun](https://www.subconscious.dev/compare/moonshot-ai-vs-stepfun.md)
- [Moonshot AI vs Runware](https://www.subconscious.dev/compare/moonshot-ai-vs-runware.md)
- [Moonshot AI vs StreamLake](https://www.subconscious.dev/compare/moonshot-ai-vs-streamlake.md)
- [Moonshot AI vs Wafer](https://www.subconscious.dev/compare/moonshot-ai-vs-wafer.md)
- [Moonshot AI vs RunInfra](https://www.subconscious.dev/compare/moonshot-ai-vs-runinfra.md)
- [Moonshot AI vs Particle.AI](https://www.subconscious.dev/compare/moonshot-ai-vs-particle-ai.md)

## Sources

- [Kimi K3 full review, Omniscient](https://www.omniscient.media/post/kimi-k3-the-full-review-the-weights-are-out-here-s-what-moonshot-didn-t-want-graded)
- [Kimi K3 guide, WhatLLM](https://whatllm.org/blog/kimi-k3)
- [Kimi K3 open weights, Nerova](https://nerova.ai/news/kimi-k3-open-weights-developer-guide)

Pricing and model lineups change often; figures are a snapshot.