b10519 b10517
ggml-org/llama.cppAug 20, 2026 → Aug 20, 20260 days between releases
Generating AI comparison
Analyzing what changed between b10519 and b10517. This page updates automatically — usually within a minute.
The Two Releases
b10519
b10519
Aug 20, 2026· 2 months ago
Refactors the server's sleep handling logic to allow access to the `/metrics` endpoint even when the server is in a sleeping state, improving observability and fixing race conditions.
Full release notesb10517
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 notes