XTU (X Transformers Unity) — a mainline-vLLM hybrid MoE plugin (CPU experts + GPU attention/long prefill). Bundles the xiaotu CPU MoE engine (AVX512-VNNI/BF16: MXFP4/FP8/INT4/BF16) so no-AMX x86 hosts (AMD EPYC) get usable CPU MoE backends through FusedMoEFactory.
Open-source, Apache-2.0 CPU implementation (BF16/FP8/WNA16/MXFP4/NVFP4) of the closed-source lk_moe CPU MoE engine, drop-in ABI-compatible with lk_moe.MOEV2
Anthropic, PBC
Visionary sponsor
Bloomberg
Visionary sponsor
Hudson River Trading
Visionary sponsor
Meta
Visionary sponsor
NVIDIA
Visionary sponsor
Microsoft
Sustainability sponsor
Depot
Continuous Integration
AWS
Cloud computing and Security Sponsor
Datadog
Monitoring
Fastly
CDN
Google
Download Analytics
Sentry
Error logging
StatusPage
Status page