Skip to content
AiBrainWorX THE CURIOUS SIDE OF AI
← Model library

Moonshot AI / RESEARCH PROFILE

Kimi-K3

A native multimodal mixture-of-experts model for long-running coding, research and knowledge work, using Kimi Delta Attention and Attention Residuals.

Source reviewed 2026-10-24 · AiBrainWorX editorial summary

Creator
Moonshot AI
Parameters
2.8T total / 104B activated
Context
1,048,576 tokens
License
Kimi K3 License

01 / CAPABILITIES

What it is built for

  • 2.8T total parameters; 104B activated
  • Native multimodal architecture
  • Always-on thinking in the published usage format

Inputs: Text + images; video understanding described by publisher

Publisher/model configuration specification. Usable context depends on deployment resources.

02 / ON YOUR MACHINE

Large server deployment

Not locally benchmarked or included as a one-click desktop download. Sparse activation does not make the full weight set fit a consumer GPU.

Launcher integration

Research profile only

Runtime and training workflow

Dedicated research runtime

Separate model-specific training workflow. Training suitability requires exact checkpoint format and prepared data; a catalog entry is not a completed training test.

Compare dataset preparation guides →

03 / WHAT IS VERIFIED

Evidence, with context

Our test status

Source-reviewed research profile; this exact checkpoint has not been benchmarked locally.

We have not added a benchmark number without documenting its test conditions. A model-card capability is not a measured speed or quality guarantee on your hardware.

04 / OWNERSHIP & USE

Kimi K3 License

Custom model license. Open weights do not imply MIT licensing or unrestricted redistribution.

AiBrainWorX authors this profile and the launcher. Moonshot AI remains the model’s creator. Models are not rebranded as our own weights.

05 / GO DEEPER

Original sources & model files

This page is our concise summary, not a mirror of the full upstream card. Review the original documentation and exact artifact before downloading.

Reviewed revision: f831ab668142 · Specifications may change after the review date.

POWER THE NEXT TEST

Help upgrade our AI test bench.

Tips help fund GPUs, RAM and storage so AiBrainWorX can test more local models, publish practical guides and build useful open tools.

Hardware donations & testing partnerships

Suggested one-time tips: $10 · $25 · $50
Choose your amount on Ko-fi.

Support is optional. Our open-source launcher stays free.