b10228

ggml-org/llama.cppb10228Aug 2, 2026by github-actions[bot]same day as b10227

What changed since b10227

medium risk

Adds DeepseekV4 multi-token-prediction and DSpark speculative decoding sidecar support, changing sidecar auto-selection priority.

New features

  • DeepseekV4 MTP (multi-token prediction) model support.
  • DSpark speculative decoding sidecars are now supported.
  • DSpark sidecars take priority over dflash during automatic sidecar selection thanks to Markov head support.

Improvements

  • Speculative decoding sidecar resolution now prefers the more capable DSpark variant when both are available.

Upgrade notes

No config or migration steps are indicated; the change is additive model/sidecar support. Be aware that auto-selection of speculative decoding sidecars now favors DSpark over dflash, so pin an explicit sidecar if you depend on the previous choice. Validate DeepseekV4 MTP and DSpark workloads before rolling out broadly.

Analyzed Nov 1, 2026

Upgrading?

Release notes summary

This release introduces support for DeepseekV4 MTP (Multi-Token Prediction) models and DSpark speculative decoding sidecars.

Key Highlights

  • Added support for DeepseekV4 MTP models.
  • Added support for DSpark speculative decoding sidecars.
  • DSpark sidecars now outrank dflash in auto-selection due to Markov head support.

Release Notes (excerpt)

<details open> DeepseekV4 MTP + DSpark (#25784) </details> **Website:** - <https://llama.app> **macOS/iOS:** - macOS Apple Silicon (arm64) - macOS Apple Silicon (arm64, KleidiAI enabled) DISABLED - macOS Intel (x64) - iOS XCFramework **Linux:** - Ubuntu x64 (CPU) - Ubuntu arm64 (CPU) - Ubuntu s390x (CPU) - Ubuntu x64 (Vulkan) - Ubuntu arm64 (Vulkan) - Ubuntu x64 (ROCm 7.2) - Ubuntu x64 (OpenVINO) - Ubuntu x64 (SYCL FP32) - Ubuntu x64 (SYCL FP16) **Android:** - Android arm64 (CPU)…

Published by ggml-org on GitHub.

Read the full release notes

Related Repositories