Month: 2026-05 — llama.cpp (ggml-org/llama.cpp) delivered targeted improvements to chat modality handling and cross-browser audio support, enhancing reliability, user experience, and browser compatibility. Key changes focus on dynamic modality detection responding to server-fetched model properties and WAV MIME support for Firefox on Linux. Fixed reactivity gaps to ensure consistent audio/video modality behavior as model properties update, enabling smoother cross-modal conversations.

Month: 2025-08 — Key bug fix in geerlingguy/linux driver addressing a PCIe L1.2 substate issue for Intel I226. Reordered device ID comparison to ensure L1.2 substate is disabled during initialization and after suspend-resume cycles, resolving a performance bottleneck that limited throughput to ~200 Mbps and enabling speeds up to ~1 Gbps when L1.2 is properly disabled. The change is captured in commit 1468c1f97cf32418e34dbb40b784ed9333b9e123 (igc: fix disabling L1.2 PCI-E link substate on I226 on init).