b11105 b11108
low risk
ggml-org/llama.cppSep 22, 2026 → Sep 22, 20260 days between releases
What's New
B11108 optimizes IQ1_M quantization performance.
New Features
- IQ1_M quantization performance improvement
Improvements
- IQ1_M prefix sum optimization
- Improved quantization speed
Upgrade Notes
No configuration changes are required; the binary is backward compatible.
The Two Releases
b11105
Pre-release
b11105
Sep 22, 2026· 30 days ago
Updates the Jinja templating engine implementation to use `const` qualifiers for statement::execute and ::visit.
Full release notesb11108
Pre-release
b11108
Sep 22, 2026· 30 days ago
Optimizes IQ1_M quantization performance by building prefix sums once per block instead of repeatedly.
Full release notes