b10506 b10505
ggml-org/llama.cppAug 20, 2026 → Aug 20, 20260 days between releases
Generating AI comparison
Analyzing what changed between b10506 and b10505. This page updates automatically — usually within a minute.
The Two Releases
b10506
b10506
Aug 20, 2026· 2 months ago
Optimizes the macOS Metal backend by implementing q8_0 dequantization using packed types, which should improve performance on Apple Silicon.
Full release notesb10505
b10505
Aug 20, 2026· 2 months ago
Introduces a new server preset option to deduplicate cached models, allowing for more efficient memory usage and cache management.
Full release notes