Skip to content

armgit / janas

Public Imported from GitHub
0

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.

NameLast commit date
..
shaders
api_aside.c
api_chat.c
api_embed.c
api_find.c
api_image.c
api_internal.h
api_keep_disk.c
api_keep.c
api_load.c
api_reply.c
api_speech_diar.c
api_speech.c
api_speech.h
api_tools.c
api.c
arch.c
arch.h
attention_scores.h
attention.c
attention.h
bf16.c
bf16.h
chat.h
constrain.c
constrain.h
deltanet.c
deltanet.h
diarize.c
expert_cache.c
expert_cache.h
generate.c
generate.h
gguf.c
gguf.h
gmatch.c
gpu_fma.c
gpu_impl.h
gpu_selfcheck.c
gpu_token_hc.c
gpu_token_impl.h
gpu_token_moe.c
gpu_token_qsa.c
gpu_token_rec.c
gpu_token.c
gpu_token.h
gpu.c
gpu.h
grammar_impl.h
grammar.c
grammar.h
iq1s_grid.h
iq2s_grid.h
iq2xs_grid.h
iq2xxs_grid.h
iq3s_grid.h
iq3xxs_grid.h
jmath.c
jmath.h
jns_write.c
jns_write.h
jns.c
jns.h
json.c
json.h
matvec.c
matvec.h
metrics.c
metrics.h
model_ana.c
model_assist.c
model_attn.c
model_ffn.c
model_gpu.c
model_hc_quant.c
model_hc.c
model_impl.h
model_janet.c
model_load.c
model_mamba.c
model_media.c
model_neurons.c
model_nh.c
model_npu.c
model_ple.c
model_qsa.c
model_rec.c
model_state.c
model_tune.c
model.c
model.h
quant_dotn.h
quant_pn.h
quant_q6k.c
quant_q6kp.h
quant.c
quant.h
sample.c
sample.h
schema_impl.h
schema_num.c
schema_re.c
schema.c
schema.h
speaker.c
speaker.h
tokenizer.c
tokenizer.h
tools.c
tools.h
tuner.c
tuner.h
unicode_data.h
vad.c
vad.h
vision_gemma_u.c
vision_gemma.c
vision_impl.h
vision_kernels.c
vision_kernels.h
vision.c
vision.h
vmath.h
whisper_align.c
whisper_mel.c
whisper_net.c
whisper_run.c
whisper.c
whisper.h