llama.cpp Releases
609 releases of ggml-org/llama.cpp
- b9594
Refactoring of vocabulary normalizer flags to improve code structure and add new options.
Jun 11, 2026
- b9592
Security update updating the vendor library LibreSSL to version 4.3.2.
Jun 10, 2026
- b9591
Optimization of MTP (Multi-Token Prediction) recurrent state handling to reduce memory overhead.
Jun 10, 2026
- b9590
Bug fix for LFM2/LFM2.5 template handler to ensure JSON schema is respected.
Jun 10, 2026
- b9553
This release improves sampler name recognition by removing a flag that restricted matching. The system now always matches canonical and alternative names in a case-insensitive manner, fixing issues with the llama-server UI.
Jun 7, 2026
- b9551
This release optimizes the key-value (KV) cache memory management by avoiding unnecessary cell copies, aiming to improve efficiency and reduce overhead.
Jun 7, 2026
- b9550
This release fixes a crash/bug where oversized assistant views caused tensor overflow during graph reservation by ensuring the shared KV cache size follows the source cache size.
Jun 7, 2026
- b9549
This release adds support for Gemma4 Multi-Token Prediction (MTP) capability to enable more efficient text generation.
Jun 7, 2026
- b9548
This release fixes a vocabulary compatibility check issue that likely ensures models load correctly across different versions or configurations.
Jun 7, 2026