37 tok/s from Qwen3-Next-80B (48 GB) on a 32 GB laptop, 24 without drafts vs 2.8 for llama.cpp; reading 62 vs 16 tok/s. Faster than llama.cpp on all 13 models compared in Q4_K_M. LLMs bigger than your RAM, pure C, CPU or GPU.
C avx2ccpu-inferenceggufinference-enginelinuxllmlocal-llmmixture-of-expertsmoenvmequantizationqwen3simdspeculative-decodingvulkan
yyjson 0.13.0, as used by Janas
- Source: https://github.com/ibireme/yyjson, tag
0.13.0, released 8 September 2026. - sha256 of the tag's tarball
(
https://github.com/ibireme/yyjson/archive/refs/tags/0.13.0.tar.gz):34e0f62a2bc11ab20d601e8ca1cc2b2079503aa45119a19133d89d19b94a0fae. - Licence: MIT (see
LICENSE). - Security: 0.13.0 carries the fix for the advisory of 8 September 2026 (integer truncation bypassing the bigint digit cap). The 2024 advisory concerns the pool allocator, which Janas does not use.
What is here
yyjson.c and yyjson.h from src/, unmodified, compiled into
janas-server without Janas's warning flags.