# llama.cpp update fixes MoE routing noise and adds Apple device distinction

> The Oct 20, 2026 release tweaks MoE router scoring and clarifies CPU vs GPU names on Apple Silicon.

Oossa · 2026-10-20 · https://oossa.com/en/llama-cpp-update-fixes-moe-routing-noise-and-adds-apple-device-distinction

The llama.cpp library released a new version on Oct 20, 2026. It changes the mixture‑of‑experts (MoE) router so the selection no longer depends on random backend noise. The fix gives each expert its own score range, reducing output differences from 4e‑3 to under 5e‑7. The same update adds the actual Apple device name (e.g., M2 Ultra) to test tables, so CPU and GPU results can be told apart.

## The facts

- Release tag b11752 published Oct 20, 2026
- MoE routing error reduced from 4e‑3 to 5e‑7

## Why it matters

Developers using llama.cpp on Apple Silicon will see more consistent model outputs and clearer performance reports.

## Sources & references

1. [ggml-org/llama.cpp b11752](https://github.com/ggml-org/llama.cpp/releases/tag/b11752) – llama.cpp, 2026-10-20

Last updated: 2026-10-20