8 min read—
Inkling 2 Small: 594B MoE Open Weights Apache 2.0
Inkling 2 Small opens its weights under Apache 2.0: a 594B MoE with 17B active and 1M context. What it takes to self-host, and how best to use it via API.
Tag: MoE
All the articles with the tag "MoE".
- 9 min read—
Colibri: Run Huge MoE Models Locally on Your PC
Colibri streams MoE experts from NVMe so huge models like GLM-5.2 and Kimi K3 run on your own PC — no datacenter GPU, no API fees, and your data stays local.