NEWS


localLLM 1.3.0 (2026-05-05)

New Functions

Bug Fixes

Backend

Bug Fixes

Behavior Changes

API Changes

Backend

Known Issues

Documentation


Performance Fix: Parallel Generation Speedup Restored

Platform Support

Backend Upgrade: llama.cpp b7825 -> b8664

New Model Support (18 new architectures)

New Chat Templates

Build System

API Compatibility

No changes to R-level API - All existing R code continues to work without modification.


localLLM 1.2.1 (2026-02-26)

Bug Fixes

localLLM 1.2.0 (2026-02-17)

Major Changes

Backend Upgrade: llama.cpp b5421 → b7825

Core Architecture Migration: KV Cache → Unified Memory API

Breaking changes in backend (transparent to R users):

Key improvements:

Batch API Modernization

Improvements

Memory Management

Error Handling

Performance

API Compatibility

No changes to R-level API - All existing R code continues to work without modification:

library(localLLM)

backend_init()
model <- model_load("model.gguf")
ctx <- context_create(model, n_ctx = 512)
result <- generate(ctx, "Hello", max_tokens = 10)
# All existing code works exactly the same

Backend Library Changes

Compilation

File Modifications

Updated files:

Unchanged:

Testing

Installation Notes

First-time Installation

install.packages("localLLM_1.2.0.tar.gz", repos = NULL, type = "source")
library(localLLM)
install_localLLM()  # Will download the new b7825 backend

Upgrading from 1.1.0

remove.packages("localLLM")
install.packages("localLLM_1.2.0.tar.gz", repos = NULL, type = "source")
library(localLLM)
install_localLLM(force = TRUE)  # Force reinstall backend

Documentation

New technical documentation:

Known Issues

Future Enhancements

Potential optimizations for future releases:

Contributors


For more information about llama.cpp, see:

localLLM 1.1.0 (2025-12-17)

Previous release notes (if any) would go here...