Skip to content

Run frontier MoE models on hardware you already own β€” pure C, zero deps, experts streamed from disk. Tiny engine, immense model. 🐦

C

Commits on main

refactor(serve): share framing with DeepSeek V4
Mario Gutierrez committed 3 months ago
8cbebce
feat(planner): OLMoE PlannerGeometry adapter (#1066)
benmaster82 committed 3 months ago
1545fac
test(v4): freeze serve wire framing
Mario Gutierrez committed 3 months ago
5fa92da
Merge pull request #1094 from JustVugg/perf/k1b-gs-idot
Vincenzo Fornaro committed 3 months ago
ad79236
Merge pull request #713 from kreuzzelg/qwen36-cuda-tier
Vincenzo Fornaro committed 3 months ago
7d1d19e
Merge pull request #1093 from JustVugg/perf/k3-lite
Vincenzo Fornaro committed 3 months ago
0d0d1ed
Merge pull request #712 from kreuzzelg/qwen36-engine
Vincenzo Fornaro committed 3 months ago
6784aa8
docs(qwen36): fix two stale references
kreuzzelg committed 3 months ago
7e226fc
e755280
refactor(serve): share framing with Kimi K3
Mario Gutierrez committed 3 months ago
cbc72fe
Merge pull request #1056 from dcutugno/pr/v4-deepgemm
Vincenzo Fornaro committed 3 months ago
cb8524b
6a9f293