Pull Requests ggerganov/llama.cpp

HIP: tune mmq/rocblas switching for RDNA4

#18816 opened 2026-01-13 16:19 by jiachengjason

CUDA: fix allignment on register spill for FA

#18815 opened 2026-01-13 16:16 by JohannesGaessler

vulkan: work around Intel ANV fp16 bug in MMQ

#18814 opened 2026-01-13 16:16 by 0cc4m

sampling : remove sampling branching in output_reserve

#18811 opened 2026-01-13 15:10 by danbev

ggml-cuda : fix warp misaligned error in debug builds Nvidia GPU ggml

#18799 opened 2026-01-13 06:36 by danbev

llama: fix integer type consistency in split helpers

#18798 opened 2026-01-13 04:15 by MaheshJakkala

CANN: fix an issue where get_env was not fully renamed devops ggml Ascend NPU

#18796 opened 2026-01-13 02:53 by noemotiovon

Unified delta net handling for Qwen3Next and Kimi Linear models model

#18792 opened 2026-01-12 19:06 by pwilkin

ci, tests : use cmake to download models and remove libcurl dependency build testing examples devops

#18791 opened 2026-01-12 18:04 by angt

server: improve slots scheduling for n_cmpl examples python server

#18789 opened 2026-01-12 17:17 by ngxson

server: fix memory reservations in populate_token_probs examples server

#18787 opened 2026-01-12 15:54 by l-austenfeld

CUDA: Factor out and re-use `block_reduce` function Nvidia GPU ggml

#18785 opened 2026-01-12 15:09 by ORippler

ggml-cpu: add RVV vec dot kernels for quantization types ggml

#18784 opened 2026-01-12 14:54 by taimur-10x

webui : send both backend_sampling == false/true examples server

#18781 opened 2026-01-12 13:22 by ggerganov

vocab: add tokenizer support for jina-embeddings-v2-base-zh python

#18756 opened 2026-01-11 11:58 by o7si

Kimi-Linear support (backend agnostic + MLA KV cache) model python ggml

#18755 opened 2026-01-11 11:55 by ymcki

fix: OOB reads in UGM tokenizer (precompiled_charsmap handling)

#18750 opened 2026-01-11 08:57 by hourhl

ggml, llama : add KV cache size limiting and block tracking infrastructure model testing examples ggml

#18747 opened 2026-01-11 00:46 by pestopoppa

fix: use actual tensor embedding dimension instead of model parameter

#18745 opened 2026-01-10 22:15 by chrismuzyn

server: add missing rerank and chat presets (#10932)

#18742 opened 2026-01-10 17:02 by ingyukoh

POC: group gate_exps and up_exps + fix mxfp4 alignment for PP boost model python

#18740 opened 2026-01-10 15:17 by am17an

llama: add canaries to Markdown files

#18735 opened 2026-01-10 11:03 by JohannesGaessler

feat: add support for WeDLM architecture python

#18731 opened 2026-01-10 02:07 by feedseawave

lookup, lookahead: fix crash when n_ctx not specified examples

#18729 opened 2026-01-10 00:09 by pestopoppa

llama: fix pooled embedding readback sizing/stride and state I/O

#18723 opened 2026-01-09 18:43 by retr0reg

model: Add VAETKI support model examples python

#18719 opened 2026-01-09 14:42 by dororodoroddo

ggml: new backend for Virglrenderer API Remoting acceleration (v2) build python ggml

#18718 opened 2026-01-09 13:29 by kpouget

Support parsing JSON into grammar for schemas with no type and no properties

#18711 opened 2026-01-09 07:37 by markrietveld

vulkan: Check maxStorageBufferRange in supports_op Vulkan ggml

#18709 opened 2026-01-09 03:40 by jeffbolznv

fix text spacing in print_info

#18708 opened 2026-01-09 02:29 by ddh0