llama.cpp changelog

TP: enable tensor split for LFM2/LFM2MOE (#26993) Assisted-by: deepseek-v4-flash Website: https://llama.app Attestations: https://github.com/ggml-org/llama.cpp/attestations/42096995 macOS/iOS: macOS Apple Silicon (arm64) macOS Apple Silicon

metal : clamp K extent in tensor API mat-mat kernel for K not a multiple of 32 (#27450) The Tensor API mat-mat path of kernel_mul_mm (GGML_METAL_HAS_TENSOR) fed a static K=32 tile to the matmul2d op on every iteration. On the last, partial

mtmd: add --mmproj-device argument (#23255) feat: add --mmproj-device arg & backwards compatible MTMD_BACKEND_DEVICE env var feat: load mmproj device backend immediately, add -mmdev shortflag fix: its a pointer now get the name clean up gen

vulkan: FA MMQ should use fp32 for Q quantization calculations (#27413) Codex found that qd could be a denorm and 1/qd would overflow. Website: https://llama.app Attestations: https://github.com/ggml-org/llama.cpp/attestations/42034841 macO

metal : dequant kv cache only for large batches (#27438) Website: https://llama.app Attestations: https://github.com/ggml-org/llama.cpp/attestations/42031082 macOS/iOS: macOS Apple Silicon (arm64) macOS Apple Silicon (arm64, KleidiAI enable

CI: Use LLVM's OpenMP over MSVC_DEBUG_non_redist on Windows (#26678) CI: Use LLVM's OpenMP over MSFT_DEBUG_non_redist on Windows Currently, we ship the non-redist debug version of microsoft's libomp. This PR changes this to official LLVM's

server: (router) lazy-load startup_models after main setup (#27424) server: (router) lazy-load startup_models after main setup only allow is_first_load to populate it nits nits 2 Website: https://llama.app Attestations: https://github.com/g

CUDA: adding switch points per HW and quant type to tune the mvq->MMQ decode crossover (#26079) CUDA: runtime GGML_CUDA_MMVQ_MAX to tune the mvq->MMQ decode crossover Add a runtime override of the mul_mat_vec_q -> MMQ batch crossover (defau

common : gracefully fallback on unsupported regex patterns in JSON schema (#26939) Website: https://llama.app Attestations: https://github.com/ggml-org/llama.cpp/attestations/42018860 macOS/iOS: macOS Apple Silicon (arm64) macOS Apple Silic

metal : dequantize quantized KV to F16 before flash attention (#27390) metal: dequantize q8_0 KV to f16 before flash attention Add a preprocessing pass for GGML_OP_FLASH_ATTN_EXT on the Metal backend: when the KV cache is quantized (Q8_0 fo

Revert "tensor-split meta backend fixes (#26502)" (#27433) This reverts commit d59d455fd8ea09e5a2e87ce2a9d668267ffb5ccd. Website: https://llama.app Attestations: https://github.com/ggml-org/llama.cpp/attestations/42012577 macOS/iOS: macOS A

ggml-cpu: gate __fp16 on __ARM_FP16_FORMAT_IEEE (#26860) ggml-cpu: gate __fp16 on __ARM_FP16_FORMAT_IEEE __ARM_NEON only signals NEON availability. The __fp16 type also needs the IEEE half format, implied on AArch64 but selected with -mfp16

server: refactor sleep handling, allow access /metrics during sleep (#27376) add cached responses refactor on_sleeping_state allow accessing metrics during sleep metrics task should not reset timer updated docs fix fix get_res_model_info ad

vulkan : dequant q8_0 KV once in coopmat1 (#25494) vulkan : dequant q8_0 KV once in coopmat1 Assisted-by: Claude (Opus 4.8) vulkan : fall back instead of aborting when FA scratch exceeds maxStorageBufferRange vulkan : require KV-cache layou

vulkan: add null checks in ggml_vk_queue_command_pools_cleanup (#27353) Guard against null queue pointers. Website: https://llama.app Attestations: https://github.com/ggml-org/llama.cpp/attestations/41873046 macOS/iOS: macOS Apple Silicon (

model : GraniteSWAForCausalLM / GraniteMoeSWAForCausalLM (#25505) feat(convert): Add conversion for GraniteSWAForCausalLM Branch: GraniteSWAForCausalLM AI-usage: full (Bob, OpenCode + Qwen3.6-35b) Signed-off-by: Gabe Goodhart [email protected]

ggml: add ggml_rope_set_offset (+ metal support) (#27120) add params cpu kernel metal kernel add test backend ops gate other backends ggml: (cuda) support ggml_rope_set_offset (#27121) rm cuda supports_op guard, fix webgpu clang-format ggml

mtmd: add mtmd_bitmap_set_mergeable (#27348) Website: https://llama.app macOS/iOS: macOS Apple Silicon (arm64) macOS Apple Silicon (arm64, KleidiAI enabled) DISABLED macOS Intel (x64) iOS XCFramework Linux: Ubuntu x64 (CPU) Ubuntu arm64 (CP

metal : dequantize q8_0 using packed types (#27370) Website: https://llama.app macOS/iOS: macOS Apple Silicon (arm64) macOS Apple Silicon (arm64, KleidiAI enabled) DISABLED macOS Intel (x64) iOS XCFramework Linux: Ubuntu x64 (CPU) Ubuntu ar

server: add dedup-cache-models preset option (#27346) Website: https://llama.app macOS/iOS: macOS Apple Silicon (arm64) macOS Apple Silicon (arm64, KleidiAI enabled) DISABLED macOS Intel (x64) iOS XCFramework Linux: Ubuntu x64 (CPU) Ubuntu

ci : add attestation for signed release artifacts (#25933) Website: https://llama.app Attestations: https://github.com/ggml-org/llama.cpp/attestations/41614541 macOS/iOS: macOS Apple Silicon (arm64) macOS Apple Silicon (arm64, KleidiAI enab

ci : Update OpenVINO to 2026.3, skip nemotron-h rollback test (#27292) update to ov-2026.3, update device drivers ci: skip nemotron-h rollback test on OpenVINO The OpenVINO backend does not support SSM_SCAN, so the Nemotron-H recurrent stat

mtmd: fix LFM2 image tiling threshold (#27057) mtmd: fix LFM2 image tiling threshold refactor testing fix fix on windows Co-authored-by: Xuan Son Nguyen [email protected] Website: https://llama.app macOS/iOS: macOS Apple Silicon (arm64) ma

sync : ggml Website: https://llama.app macOS/iOS: macOS Apple Silicon (arm64) macOS Apple Silicon (arm64, KleidiAI enabled) DISABLED macOS Intel (x64) iOS XCFramework Linux: Ubuntu x64 (CPU) Ubuntu arm64 (CPU) Ubuntu s390x (CPU) Ubuntu x64

build : fix xcframework + cmake clean-up (#27304) xcframework : fix build mtmd : remove unused include path vendor : use vendor::hash alias target in cmake CMake reserves "::" in target names for imported/alias targets, so the real target k

cuda : skip UMA override for HIP builds (#27083) AMD APUs report accurate memory via hipMemGetInfo. Using MemAvailable over-promises on small-carveout systems. fixes #18159 Website: https://llama.app macOS/iOS: macOS Apple Silicon (arm64) m

ci : push release tag explicitly in release.yml (#27261) Add a "Create and push git tag" step to the release job, right before the "Create release" step. The tag is created with git tag and pushed with the deploy key already configured by t

sycl: fix thread/block count in quantized cpy kernel launches (#27160) Adjusts the thread/block count to be proportional to the size of the quant, reducing under/over subscription. Largest perf improvement is the q4_0 -> f32 path, with, on

[SYCL] support OP OPT_STEP_ADAMW, OPT_STEP_SGD (#25268) fix conflict fix conflict of ops.md fix conflict of ops.md update the ops.md Co-authored-by: Neo Zhang Jianyu [email protected] Website: https://llama.app macOS/iOS: macOS Apple S

model : remove some ggml_concat (#27176) Co-authored-by: Xuan Son Nguyen [email protected] Website: https://llama.app macOS/iOS: macOS Apple Silicon (arm64) macOS Apple Silicon (arm64, KleidiAI enabled) DISABLED macOS Intel (x64) iOS XCFra

chat: refactor handling supports_string_content / supports_typed_content (#27130) better supports_string_content cap detect test: add "skip" messages_inp_normalizer Website: https://llama.app macOS/iOS: macOS Apple Silicon (arm64) macOS App

llama : check LoRA tensor data is within file bounds (#27056) llama : check LoRA tensor data is within file bounds Update src/llama-adapter.cpp Co-authored-by: Sigbjørn Skjæret [email protected] Co-authored-by: Sigbjørn Skjær

model: add Kimi-K3 text model (#26185) model: add Kimi-K3 text model Hybrid KDA (linear) + MLA (full) attention as in Kimi-Linear-48B, plus five things that architecture does not have: cross-layer residual attention (attn_res_block_size) la

server: re-design yield_to_queue thread model (#27133) run common_speculative_process in worker swap worker <--> main thread design Website: https://llama.app macOS/iOS: macOS Apple Silicon (arm64) macOS Apple Silicon (arm64, KleidiAI enabl

vendor: update BoringSSL to 0.20260813.0 (#27099) Website: https://llama.app macOS/iOS: macOS Apple Silicon (arm64) macOS Apple Silicon (arm64, KleidiAI enabled) DISABLED macOS Intel (x64) iOS XCFramework Linux: Ubuntu x64 (CPU) Ubuntu arm6

common: support --models-dir loading MTP assistant models (#24431) common: support --models-dir loading MTP assistant models common: preset: check for MTP models with strict prefix common: preset: Take advantage of PR #27005 handle other dr

fix: check gguf array type before reading (#27075) fix: check gguf array type before reading update skill Website: https://llama.app macOS/iOS: macOS Apple Silicon (arm64) macOS Apple Silicon (arm64, KleidiAI enabled) DISABLED macOS Intel (

vulkan: add SHMEM_STRIDE_PAD/APPLY_SLM_A_RESHAPE for coopmat1 on Intel Xe (#25380) vulkan: add SHMEM_STRIDE_PAD/APPLY_SLM_A_RESHAPE for coopmat mul_mm on Intel Xe vulkan: fix shmem estimate for Intel SHMEM_STRIDE_PAD=0 in matmul_shmem_suppo

common: migrate the deprecated --mmap/--no-mmap to --load-mode (#26934) Replace the deprecated --mmap, --no-mmap, --mlock, and --direct-io flags with the unified --load-mode argument across scripts, examples, and documentation. Internal war

model : add support for MiniMaxText01ForCausalLM and MiniMaxM1ForCausalLM (#27018) llama : support for MiniMax-Text-01 model chore : renames to match the other MiniMax models model : add logits mask as MiniMax-Text-01 embeddings tensor has

mtmd, common: various fixes (#27071) apply fixes cont revert gguf fix Website: https://llama.app macOS/iOS: macOS Apple Silicon (arm64) macOS Apple Silicon (arm64, KleidiAI enabled) DISABLED macOS Intel (x64) iOS XCFramework Linux: Ubuntu x

jinja : fix quadratic cost in gather_string_parts (#27034) jinja : fix quadratic cost in gather_string_parts fix some comments remove test Website: https://llama.app macOS/iOS: macOS Apple Silicon (arm64) macOS Apple Silicon (arm64, KleidiA

chat : pass reasoning_effort to template chat: add reasoning_effort to common_chat_templates_inputs Store OpenAI Chat Completions reasoning_effort and make it available to jinja templates (with model specific translations where required). A

sync : ggml Website: https://llama.app macOS/iOS: macOS Apple Silicon (arm64) macOS Apple Silicon (arm64, KleidiAI enabled) DISABLED macOS Intel (x64) iOS XCFramework Linux: Ubuntu x64 (CPU) Ubuntu arm64 (CPU) Ubuntu s390x (CPU) Ubuntu x64

ggml : recurrent state rollback for ggml_ssm_scan (#26623) Initial changes for Recurrent state rollback for nemotron for cpu and cuda Removing CPU RS rollback. Will enable it in subsequent PRs addition of test case Removing assert and calli

llama : allow virtual igpu devices (#26953) llama : allow virtual igpu devices cont : better comment Website: https://llama.app macOS/iOS: macOS Apple Silicon (arm64) macOS Apple Silicon (arm64, KleidiAI enabled) DISABLED macOS Intel (x64)

server: allow accessing /metrics and /slots during llama_decode() (#27041) server_queue::worker call llama_decode inside yield_to_queue also handle process_mtmd_chunk clean up nits rm test Website: https://llama.app macOS/iOS: macOS Apple S

tests : replace personal home directory paths with generic placeholders (#27043) Scrub developer-specific /home// paths from example docs and test fixtures so they don't leak into the tree. examples/test-cmake/README.md: /home/danbev/... ->

sycl: fuse mul_mat(gate) + mul_mat(up) + GLU for q4_K dense FFN (#26779) Measured on Arc Pro B70 (Battlemage, Level Zero), llama-bench -r 20, two interleaved rounds, tg128: qwen2.5-3B-Instruct Q4_K_M 154.18 -> 158.53 t/s +2.8% gemma-2-2b-it

ggml: force single thread on wasi (#25686) Website: https://llama.app macOS/iOS: macOS Apple Silicon (arm64) macOS Apple Silicon (arm64, KleidiAI enabled) DISABLED macOS Intel (x64) iOS XCFramework Linux: Ubuntu x64 (CPU) Ubuntu arm64 (CPU)

sycl: fuse the gated-delta-net state writeback cpy (#26643) Port of https://github.com/ggml-org/llama.cpp/pull/23940. Arc Pro B70, Qwen 3.6 27B Q4_K - Medium (48 of its 64 blocks run gated_delta_net), -ngl 99 -fa 1 -ctk f16 -ctv f16 -b 2048

dflash : clarify output logging of target_layer_ids (#27013) This commit tries to make the logging of target_layer_ids a bit clearer and easier to read. Currently the output generated looks like this: 0.00.468.624 D load_arch_hparams: DFlas

common: apply CPU parameters across tools (#27026) Website: https://llama.app macOS/iOS: macOS Apple Silicon (arm64) macOS Apple Silicon (arm64, KleidiAI enabled) DISABLED macOS Intel (x64) iOS XCFramework Linux: Ubuntu x64 (CPU) Ubuntu arm

OpenVINO: Qwen3.5, memory optimization, and test-recurrent-state-rollback (#26952) OpenVINO backend: 1) enable gpt-oss moe on OV bk; 2) enable mxfp4 support OpenVINO backend: disable TOPK_MOE op test OpenVINO Backend: Add op FILL support Op

[SYCL] Support host pinned mem to improve SYCL Host-to-Device Memory Access (#26789) support host pinned mem, ggml_backend_sycl_host_buffer_type_get_max_size, fix the thread-safe issue Website: https://llama.app macOS/iOS: macOS Apple Silic

chat : fix LFM2 tool call arg name prefix ambiguity (#26960) Assisted-by: Claude Opus 5 Website: https://llama.app macOS/iOS: macOS Apple Silicon (arm64) macOS Apple Silicon (arm64, KleidiAI enabled) DISABLED macOS Intel (x64) iOS XCFramewo

server : serve index.html with no-cache (#27006) index.html was served with max-age=31536000, immutable like the hashed assets, but its name is stable while its contents change every build, so a cached copy pins the UI to an old build. It n

spec : auto-detect mtp draft model type (#27005) Website: https://llama.app macOS/iOS: macOS Apple Silicon (arm64) macOS Apple Silicon (arm64, KleidiAI enabled) DISABLED macOS Intel (x64) iOS XCFramework Linux: Ubuntu x64 (CPU) Ubuntu arm64