← Catalogsource ↗
Moonlight-16B-A3B-Instruct
AvailableMoonshot AIOpen source
MIT-licensed 16B/3B-active MoE trained with Moonshot's scalable Muon optimizer experiments.
Specifications
- License
- Open source · MIT
- Weights
- Downloadable
- Architecture
- Mixture-of-Experts
- Parameters
- 16B · 3B active
- Context window
- 8K tokens
- Max output
- —
- Knowledge cutoff
- —
- Price (in / out, $/M)
- —
- Modalities
- TextCode
Benchmarks
MMLUself-reported
70%Vendor-reported figures are claims until independently verified. See methodology.