Qeravio
Canonical AI event

CUDA graph usage improved for MTP in llama.cpp draft

CUDA graph improvements for MTP draft include support for Ubuntu x64 with CUDA 12 and 13.

17 Sept 20261 verified claims1 sources1 observations
What happened

The official source reports this update: b11007: Enable CUDA graph for MTP draft.

Why it matters

This official update documents a development concerning b11007: Enable CUDA graph for MTP draft. Its practical significance depends on the scope and evidence stated by the source.

What to watch next

Read the official source update and verify its stated scope, evidence and timing before acting on it.

Connected knowledge

Entities affected by this event

Llama Cppsubject
Continue this topic
infrastructureFused QKV gemma4, qwen35 models now calculate split states for attention.17 Sept 2026infrastructureHexagon adds support for K-Quants Q4_K and Q6_K kernels17 Sept 2026infrastructureLlama app now supports zeroed rope probe in decoding graph.16 Sept 2026infrastructureLlama.cpp improves im2col access patterns in CUDA/HIP16 Sept 2026infrastructurespacemit fixes int16 data transpose bug in permute_transpose_impl16 Sept 2026infrastructureServer discards cached graphs on buffer free to prevent remote code execution.16 Sept 2026
Evidence trail

Sources behind the event

llama.cpp official releasesb11007: Enable CUDA graph for MTP draft
Editorial presentation

Read the public story

Open story