Pull Requests ggml-org/llama.cpp

reasoning custom sampling params testing

#25699 by nmarasoiu was closed 2026-07-15 05:14

DeepseekV4: support bounded partial rollback testing

#25691 by YanissAmz was closed 2026-07-15 02:02

llama: batch the device reads when saving a state

#25687 by YanissAmz was closed 2026-07-15 02:02

ggml: force single thread on wasi ggml

#25685 by MendyBerger was closed 2026-07-14 19:49

model: add DSpark speculator support for DeepSeek-V4 targets documentation model server conversion

#25683 by YanissAmz was closed 2026-07-15 02:02

server: draft acceptance stats were wrong whenever the target could n… server

#25682 by YanissAmz was closed 2026-07-15 02:04

Enhance ggml_cuda_mul_mat_cublas with compute type fallback ggml CUDA

#25680 by theIvanR was closed 2026-07-15 04:59

hexagon: fix HMX queue signal enum narrowing issue ggml Hexagon

#25677 by chyan8 was merged 2026-07-14 19:27

opencl: do not call OpenCL 3.0 API functions for older drivers ggml OpenCL

#25673 by wanghqc was merged 2026-07-15 02:53

opencl: avoid the vec path in GEMV for unaligned row stride ggml OpenCL

#25671 by wanghqc was merged 2026-07-14 19:27

Infcore documentation nix examples devops

#25662 by Nasferatuss was closed 2026-07-14 14:04

server: add --cors-* options server

#25655 by ngxson was merged 2026-07-14 15:23

metal: enable F16 support for leaky ReLU ggml Apple Metal

#25651 by kjg0724 was closed 2026-07-14 09:09

ggml : add a set of functions for checking contiguity of inner tensor dimensions ggml

#25650 by fairydreaming was merged 2026-07-14 12:37

server : refactor prompt cache state ownership server

#25649 by ggerganov was merged 2026-07-14 15:25

server : fix checkpoints not being freed when a slot's prompt is cleared testing server

#25647 by AlejandroParedesLT was closed 2026-07-14 07:36

deepseek4: NextN/MTP speculative decoding (converter + graph) documentation model conversion

#25642 by bownux was closed 2026-07-14 02:42

convert_hf_to_gguf.py: Refactor how `--mtp` / `--no-mtp` is enabled for a model conversion

#25641 by tarruda was merged 2026-07-14 09:43

opencl: handle OOB write in noshuffle GEMV kernels (odd ne01) testing ggml OpenCL

#25640 by wanghqc was merged 2026-07-14 20:46

opencl: fix a dp4a bug for devices where cl_khr_integer_dot_product is unavailable ggml OpenCL

#25639 by wanghqc was merged 2026-07-14 15:08

Fix missing space in tool-call prompt request server/ui

#25634 by bill88t was merged 2026-07-14 15:23

CUDA: add GGML_OP_LIGHTNING_INDEXER kernel (F16/F32 K) ggml CUDA

#25632 by Dorijan10 was closed 2026-07-14 02:43

ui: fix mcp panel for toggle + timeout + proxy + ON/OFF state server/ui

#25631 by ServeurpersoCom was merged 2026-07-14 14:50

Re-add the llama-proxy toggle in the "Add New Server" menu & fix the missing space in tool-call requests server/ui

#25630 by bill88t was closed 2026-07-13 18:03

readme : add link to maintainer PRs documentation

#25621 by ggerganov was merged 2026-07-13 13:07

tests: export-graph-ops: exit gracefully when called w/o arguments testing

#25619 by ckastner was merged 2026-07-14 10:15

tests: Harmonize header use testing merge ready

#25616 by ckastner was merged 2026-07-13 12:36

ppu: dlopen .so seam for FA / MoE-GEMM / gated-delta-net CUDA kernels documentation testing ggml CUDA

#25607 by DrXuQian was closed 2026-07-13 01:59

Fix Minimax 2 EAGLE3 support model merge ready

#25604 by adrianisk was merged 2026-07-13 13:22

Webui bugfix for single-model mode usage server/ui

#25599 by intel00000 was closed 2026-07-14 09:45