b10782 b11104
low risk
ggml-org/llama.cppSep 3, 2026 → Sep 22, 202619 days between releases
What's New
Adds multi-address binding support to llama-server and rejects port 0 with multiple addresses
Breaking Changes
- Rejects --port 0 when multiple TCP addresses are specified
New Features
- Supports binding llama-server to multiple TCP addresses simultaneously
- Improved argument handling
Upgrade Notes
You can now bind llama-server to multiple TCP addresses simultaneously. Note that using --port 0 is now rejected when multiple addresses are provided.
The Two Releases
b10782
Pre-release
b10782
Sep 3, 2026· 2 months ago
Fixes a CUDA multi-GPU issue where graph optimization was skipped due to concurrent stream handling. The fix allows concurrent streams per split and explicitly sets the device to prevent CUDA events from sticking to the wrong GPU.
Full release notesb11104
Pre-release
b11104
Sep 22, 2026· 30 days ago
Adds the ability for the llama-server to bind to multiple TCP addresses simultaneously, improving network flexibility.
Full release notes