llama.cpp Releases
684 releases of ggml-org/llama.cpp
- b9637
Added a dedicated chat parser for the Cohere2-MoE (North Code) model format to improve chat interaction handling.
Jun 14, 2026
- b9632
Enhanced the Jinja template engine by adding aliases for the count, d, and e filters.
Jun 14, 2026
- v1.4.0
This release adds support for new AI models like Gemma-4 and LiquidAI, introduces automatic model reproduction via JSON files, adds plain text prompt datasets, and improves UI consistency and error handling.
Jun 14, 2026
- b9631
Fixed a bug in the CLI where preserved tokens were not being copied correctly during processing.
Jun 14, 2026
- b9630
Updated the TINY_AYA tokenizer vocabulary to include support for the Cohere2-MoE architecture.
Jun 14, 2026
- b9628
Added SYCL (ComputeCpp) support to the release check scripts to ensure consistency across builds.
Jun 14, 2026
- b9627
Fixed a crash in the llama-ui-embed feature that occurred when no asset directory was provided.
Jun 13, 2026
- b9626
Major update adding support for the Cohere2-MoE (Mixture of Experts) architecture, including specific tensor handling, tokenizer updates, and fixes for MTP and sliding windows.
Jun 13, 2026
- b9625
Fixed a critical bug in the Jinja template engine where negative step slicing with start/stop values was not working correctly.
Jun 13, 2026
- b9624
Updated the UI build process to use build-time gzip compression to reduce asset sizes while preserving original filenames and paths.
Jun 13, 2026
- b9623
A bug fix release addressing issues with the Jinja template processing library, specifically regarding the split and replace functions, and reserve size.
Jun 13, 2026
- b9622
Enhances the Vulkan backend to support non-contiguous memory layouts for unary and GLU operations. Includes optimizations for index calculation and support for expm1.
Jun 13, 2026
- b9621
Fixes a UI issue where file names and paths were not preserved correctly.
Jun 13, 2026
- b9620
Improves the server's static asset handling by bundling UI assets into an archive and standardizing file naming. Includes code cleanup and prettier formatting.
Jun 13, 2026
- b9616
Fixes CI release process issues to ensure builds succeed. This involves removing a missing test and addressing other release pipeline failures.
Jun 12, 2026
- b9611
Refactors the fit module to avoid unnecessary includes of llama-ext.h in fit.h.
Jun 12, 2026
- b9610
A general synchronization commit for the ggml library.
Jun 12, 2026
- b9608
Updates the embedded cpp-httplib library to version 0.47.0. This is a vendor update with no functional changes to the core library logic.
Jun 12, 2026
- b9606
Implements support for EAGLE3 speculative decoding, enabling layer input extraction for improved generation efficiency. This includes compatibility with Gemma4 eagle3 and fixes for parameter bugs and multi-sequence issues.
Jun 12, 2026
- b9605
Adds support for concatenating tensors of scalar types in the CUDA backend. Also fixes a Metal CI issue during the update.
Jun 12, 2026