Local Ai
b9163
b9163 is a llama.cpp release that adds a custom Unicode regex handler for Qwen3.5's tokenizer to prevent stack overflows on long inputs . The release also includes improvements to SYCL memory manageme
b9163 is a llama.cpp release that adds a custom Unicode regex handler for Qwen3.5's tokenizer to prevent stack overflows on long inputs . The release also includes improvements to SYCL memory management by consolidating device allocation/deallocation functions and refactoring SYCL_CHECK error wrapping patterns . This build represents ongoing optimization work in the llama.cpp C/C++ inference engine for language model execution.
Source: llama.cpp Releases | 2026-05-15