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 */
/*
* evals_metrics.h - the text similarity metrics of the evals' graders
* (evals_metrics.c).
*/
#ifndef JANAS_SERVER_EVALS_METRICS_H
#define JANAS_SERVER_EVALS_METRICS_H
/* The score, 0 to 1, of input against reference by metric: cosine,
fuzzy_match, bleu, gleu, meteor, rouge_1 to rouge_5, rouge_l. */
double srv_eval_similarity(const char *metric, const char *input,
const char *reference);
/* 1 when metric is one of them. */
int srv_eval_metric_ok(const char *metric);
#endif