Pull Requests ggerganov/llama.cpp

ui: take enabled frontend tools into account server/ui

#25368 by Cr4xy was closed 2026-07-07 14:40

cuda : add support for f16->f16 GGML_OP_SET_ROWS ggml merge ready CUDA

#25367 by fairydreaming was merged 2026-07-08 11:24

opencl: general flash attention decode performance optimizations ggml OpenCL

#25366 by wanghqc was merged 2026-07-07 02:57

model: add support for Hy3 (HYV3ForCausalLM) model conversion

#25364 by YanissAmz was closed 2026-07-06 18:37

server: fix deadlock in load_models() when erasing a finished download server

#25358 by ServeurpersoCom was merged 2026-07-06 17:26

server: temporary skip model downloading API test server

#25355 by ngxson was merged 2026-07-06 14:10

vulkan : check src0 type in GGML_OP_SET_ROWS to avoid failures due to unimplemented f16 support Vulkan ggml

#25351 by fairydreaming was merged 2026-07-07 04:56

server : add timings and progress to responses API stream server

#25348 by thom-dev-fr was merged 2026-07-07 14:13

ggml : add support for CPU f16->f16 GGML_OP_SET_ROWS testing ggml

#25344 by fairydreaming was merged 2026-07-08 03:46

ui: Context usage gauge and panel server/ui

#25340 by allozaur was merged 2026-07-08 07:22

ggml-cpu: use UE4M3 LUT in ARM NVFP4 dot product ggml merge ready

#25331 by ragz4125 was merged 2026-07-06 11:06

server: fan out router metrics by model server

#25330 by tobi was closed 2026-07-05 19:09

server: expose model resource metrics server

#25329 by tobi was closed 2026-07-05 19:09

server: expose prompt cache metrics server

#25328 by tobi was closed 2026-07-05 19:09

server: expose speculative decoding metrics server

#25327 by tobi was closed 2026-07-05 19:09

server: evict LRU resident model on load failure (VRAM pressure) server

#25326 by tobi was closed 2026-07-05 19:12

server: eagerly create prompts log directory at startup documentation server

#25322 by rankaiyx was merged 2026-07-08 07:45

cuda, cpu : fix GQA head mapping in GDN kernel ggml CUDA

#25315 by Bmasele was closed 2026-07-05 10:52

ggml-cuda : optimize conv_transpose_1d indexing ggml CUDA

#25310 by adavyas was merged 2026-07-06 03:49

ui: restore Ctrl+B sidebar toggle shortcut server/ui

#25307 by ServeurpersoCom was merged 2026-07-06 08:30

server-context: chain the caller-provided load_progress_callback instead of discarding it server

#25306 by mtmcp was closed 2026-07-05 09:25

cuda : concat implementation for quantized types ggml merge ready CUDA

#25303 by fairydreaming was merged 2026-07-05 15:26

ui: fake 200 for proxy DELETE req server/ui

#25298 by ngxson was merged 2026-07-06 06:41

server: handle DELETE on /cors-proxy for MCP streamable-http server

#25296 by tlemaire was closed 2026-07-04 09:16

scripts : use HF_TOKEN when downloading UI assets

#25280 by angt was merged 2026-07-06 07:53

llama-batch: enable parallel sequences for partial rollback model

#25278 by am17an was merged 2026-07-08 07:55

ggml-backend-meta: abort if we see a multi buffer ggml

#25276 by netrunnereve was merged 2026-07-05 18:38

vulkan: implement PAD_REFLECT_1D documentation Vulkan ggml

#25272 by oliver0006 was closed 2026-07-03 16:45

cuda: enable topk-moe fusion for 288 experts testing ggml CUDA

#25267 by pwilkin was merged 2026-07-03 13:36

server: also free backend on sleep server

#25265 by ngxson was closed 2026-07-03 14:51