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
/* SPDX-License-Identifier: GPL-3.0-or-later
Copyright (C) 2026 Maurizio Cammalleri */
/*
* stop.h - a program that runs until it is asked to stop: Ctrl+C, a
* termination (SIGTERM), on Windows the console closed or the user
* logging off. The request is awaited by one thread; the others never
* see it.
*/
#ifndef JANAS_COMMON_STOP_H
#define JANAS_COMMON_STOP_H
/* Before any thread is made: the requests to stop held back from every
thread but the one that awaits them; a write to a closed connection an
error, not the end of the process (SIGPIPE ignored). */
void janas_stop_prepare(void);
/* Waits for a request to stop. */
void janas_stop_wait(void);
#endif