Moonshot AI Releases 2.8T Parameter Kimi K3 Weights with Custom Licensing Terms
Moonshot AI has made the weights for its 2.8 trillion parameter Kimi K3 model publicly available on Hugging Face as a 1.56TB download. The release features a customized license requiring separate commercial agreements for large Model-as-a-Service providers and mandatory UI attribution for high-revenue apps.
Impact: Medium
Why it matters
Developers planning self-hosted or commercial API deployments of Kimi K3 must audit their revenue and distribution channels against Moonshot's custom licensing constraints.
TL;DR
- 01Kimi K3 weights (1.56TB) are hosted on Hugging Face for self-hosting setups.
- 02Commercial MaaS entities with >$20M revenue require custom licensing agreements.
- 03OpenRouter serves K3 via 7 providers at $3/M input and $15/M output tokens.
Key facts
- Model Parameters
- 2.8 Trillion
- Hugging Face Download Size
- 1.56 TB
- OpenRouter Input Price
- $3.00 / 1M tokens
- OpenRouter Output Price
- $15.00 / 1M tokens
- MaaS Revenue Threshold
- $20M aggregate over 12 months
Model Weights and Hardware Requirements
Moonshot AI has uploaded the full weights for the 2.8 trillion parameter Kimi-K3 model to Hugging Face. The full repository weighs in at 1.56TB, requiring multi-GPU server nodes or high-throughput cluster storage to mirror and serve locally.
Custom Licensing Constraints
Moonshot explicitly avoids labeling the release as open source, adopting the term 'open weight'. Key legal conditions include:
- Model as a Service (MaaS): Organizations generating over $20 million in aggregate revenue over a consecutive 12-month period must execute a formal separate agreement with Moonshot AI before commercial use.
- UI Attribution: Products serving over 100 million monthly active users (MAU) or earning $20 million monthly revenue must prominently display
Kimi K2in the product interface.
API Pricing and Provider Access
For developers preferring managed endpoints over running 1.56TB local weights, OpenRouter currently offers access across seven hosting providers. Standard pricing is aligned at $3.00 per 1M input tokens and $15.00 per 1M output tokens.
Try it in 2 minutes
pip install huggingface_hub
huggingface-cli download moonshotai/Kimi-K3 --local-dir ./kimi-k3bash
✓ When to use
- Evaluating frontier open-weight models for self-hosted enterprise infrastructure.
- Accessing 2.8T MoE reasoning via OpenRouter API without hosting hardware.
✕ When NOT to use
- Building a commercial MaaS backend exceeding $20M revenue without custom Moonshot contracts.
- Deploying on low-bandwidth or edge servers unable to store 1.56TB weights.
What to do today
- Audit commercial revenue models before deploying Kimi K3 in production MaaS offerings.
- Evaluate OpenRouter API endpoints if local infrastructure cannot store 1.56TB of weights.
Sources