Visual TL;DR
llamafile v0.10.5 releasedCore
new version of llamafile ships with updated core llama.cpp integration
From the article 7 mentionsThe latest release of llamafile, version 0.10.5, is now available, bringing enhanced support for running substantial AI models directly on personal hardware.
Outdated llama.cppDriver
older llamafile builds lacked support for new model architectures and quantization methods
From the article 4 mentionsThese models, previously available as GGUF weights, were inaccessible in older llamafile builds due to outdated llama.cpp support.
Updated llama.cpp coreCore
integrates recent llama.cpp changes, enabling support for new model types
From the article 4 mentionsThis update tracks a more recent llama.cpp core, unlocking the ability to deploy models like the 6GB Ternary Bonsai 27B and the 118B parameter Laguna-S-2.1 coding MoE.
Packaging, docs upgradesContext
full details on announcement, including improved packaging and documentation
Support for big modelsEffect
now runs Ternary Bonsai 27B and Laguna-S-2.1 directly on local hardware
From the article 3 mentionsBeyond model support, llamafile v0.10.5 also introduces useful enhancements.
Ternary Bonsai 27BContext
6GB compressed Qwen3.6-27B model using ternary weights for efficiency
From the article 2 mentionsThe inclusion of Ternary Bonsai 27B is particularly noteworthy.
Laguna-S-2.1 MoEContext
118B parameter coding MoE model now accessible on consumer machines
From the article 3 mentionsThis update tracks a more recent llama.cpp core, unlocking the ability to deploy models like the 6GB Ternary Bonsai 27B and the 118B parameter Laguna-S-2.1 coding MoE.
Efficient local AIOutcome
previously unwieldy models now fit and run efficiently on consumer-grade machines
From the articleThis focus on efficient packaging and integration is central to llamafile's value proposition.