b10517 b10516
ggml-org/llama.cppAug 20, 2026 → Aug 20, 20260 days between releases
Generating AI comparison
Analyzing what changed between b10517 and b10516. This page updates automatically — usually within a minute.
The Two Releases
b10517
b10517
Aug 20, 2026· 2 months ago
Optimizes Flash Attention in the Vulkan backend by dequantizing KV cache only once in cooperative matrix operations and handling edge cases like FA scratch buffer limits.
Full release notesb10516
b10516
Aug 20, 2026· 2 months ago
Addresses a potential crash in the Vulkan backend by adding null checks in the queue command pools cleanup function.
Full release notes