| 2026-07-24 |
Pascal | ui: fix MCP server display name conflicts in tools... |
commit | commitdiff | tree |
| 2026-07-24 |
Nigel Bosch | server: support "reasoning_effort": "none" in OAI API... |
commit | commitdiff | tree |
| 2026-07-24 |
Xuan-Son Nguyen | llama: various bug fixes (#26051) |
commit | commitdiff | tree |
| 2026-07-24 |
Johannes Gäßler | HIP: remove rocWMMA FlashAttention (#26046) |
commit | commitdiff | tree |
| 2026-07-24 |
Hongqiang Wang | opencl: cache compiled cl_program binaries on disk... |
commit | commitdiff | tree |
| 2026-07-24 |
kumaal | opencl: do not treat NULL-mask flash attention as causa... |
commit | commitdiff | tree |
| 2026-07-24 |
Xuan-Son Nguyen | skill: create `add-new-model` and `code-review` (#26042) |
commit | commitdiff | tree |
| 2026-07-24 |
Pascal | UI: Fix settings precedence, Factory < Admin (--ui... |
commit | commitdiff | tree |
| 2026-07-24 |
Matt Thompson | cohere2 moe template parser: enforce JSON schema for... |
commit | commitdiff | tree |
| 2026-07-24 |
Xuan-Son Nguyen | vendor: update subprocess.h (#26061) |
commit | commitdiff | tree |
| 2026-07-24 |
Max Krasnyansky | hexagon: further improved pipeline of the core bits... |
commit | commitdiff | tree |
| 2026-07-23 |
adgup-qti | hexagon: fix Windows crash when op_poll is enabled... |
commit | commitdiff | tree |
| 2026-07-23 |
Johannes Gäßler | CUDA: fix external compilation of q1_0 MMQ (#25778) |
commit | commitdiff | tree |
| 2026-07-23 |
Aaron Teo | args: refactor mlock/mmap/directio into load-mode ... |
commit | commitdiff | tree |
| 2026-07-23 |
Pascal | contrib: fix leftovers from the AI usage policy update... |
commit | commitdiff | tree |
| 2026-07-23 |
Ilia Ilmer | metal : add f16 type support to leaky relu (#25981) |
commit | commitdiff | tree |
| 2026-07-23 |
Shahir BIn... | conversion: fix non-MoE NomicBert GGUF conversion error... |
commit | commitdiff | tree |
| 2026-07-22 |
Xuan-Son Nguyen | contrib: allow all AI-generated code in general (#26012) |
commit | commitdiff | tree |
| 2026-07-22 |
Pascal | ui: Add a "Default" option for the reasoning selector... |
commit | commitdiff | tree |
| 2026-07-22 |
Oliver Simons | CUDA: Improve NVFP4 W4A4 activation quantization (... |
commit | commitdiff | tree |
| 2026-07-22 |
Todor Boinovski | hexagon: activation ops update (#25974) |
commit | commitdiff | tree |
| 2026-07-22 |
Niklas Wenzel | mtmd: use RAII for setting and resetting non-causal... |
commit | commitdiff | tree |
| 2026-07-22 |
rankaiyx | feat(ui): add symbolic math support to JS sandbox via... |
commit | commitdiff | tree |
| 2026-07-22 |
Piotr Wilkin... | minor: fix reasoning preserve var for DS4 [no ci] ... |
commit | commitdiff | tree |
| 2026-07-22 |
Pascal | common: infer the speculative type from the draft repo... |
commit | commitdiff | tree |
| 2026-07-22 |
Piotr Wilkin... | Fix DeepSeek4 crafted template (#25414) |
commit | commitdiff | tree |
| 2026-07-22 |
shalinib-ibm | ggml: enable PowerPC backend variants on AIX (#25983) |
commit | commitdiff | tree |
| 2026-07-22 |
KyleHagy | ci : fix SYCL package shared library lookup (#25987) |
commit | commitdiff | tree |
| 2026-07-22 |
m1el | webgpu : add CONV_2D_DW (depthwise conv2d) kernel ... |
commit | commitdiff | tree |
| 2026-07-22 |
Pascal | cuda: GET_ROWS quants (#25962) |
commit | commitdiff | tree |
| 2026-07-22 |
helanfxz | llama-arch: fix DeepSeek4 APE tensor op (#25945) |
commit | commitdiff | tree |
| 2026-07-22 |
Joe Rowell | Add support for Laguna XS.2 & M.1 (#25165) |
commit | commitdiff | tree |
| 2026-07-21 |
wendadawen | convert: fix handle HunyuanVL XD-RoPE config (#25514) |
commit | commitdiff | tree |
| 2026-07-21 |
Gerben van V | mtmd : use align_corners for qwen3vl vision position... |
commit | commitdiff | tree |
| 2026-07-21 |
Wei Wang | hexagon: check tensor type when reusing descriptors... |
commit | commitdiff | tree |
| 2026-07-21 |
Aman Gupta | cuda: add sqrt_softplus in topk-moe for dsv4 (#25896) |
commit | commitdiff | tree |
| 2026-07-21 |
Kamalesh VS | kleidiai : warn once when a weight type has no KleidiAI... |
commit | commitdiff | tree |
| 2026-07-21 |
Pascal | common: resolve draft repo to its requested sidecar... |
commit | commitdiff | tree |
| 2026-07-21 |
Pascal | server: return 400 instead of 500 on validation error... |
commit | commitdiff | tree |
| 2026-07-21 |
fairydreaming | server : properly handle null llama_context (#25868) |
commit | commitdiff | tree |
| 2026-07-21 |
Winston Ma | vulkan: Refactor vk_queue to use per-instance mutexes... |
commit | commitdiff | tree |
| 2026-07-21 |
Markus Ebner | ggml-openvino: Add GGML_BACKEND_DL_IMPL invocation... |
commit | commitdiff | tree |
| 2026-07-21 |
Piotr Wilkin... | CUDA: vectorize same-type get_rows with int4 copy ... |
commit | commitdiff | tree |
| 2026-07-20 |
Todor Boinovski | hexagon: add CLAMP op (#25934) |
commit | commitdiff | tree |
| 2026-07-20 |
Aleksander... | ui: Sidebar Conversations Bulk Action + Improved Settin... |
commit | commitdiff | tree |
| 2026-07-20 |
Aman Gupta | llama_dsv4: write only used rows in state (#25325) |
commit | commitdiff | tree |
| 2026-07-20 |
Pascal | ui: fix collapsed user bubble with markdown rendering... |
commit | commitdiff | tree |
| 2026-07-20 |
Pascal | UI: fix Settings/Display tool call content toggle ... |
commit | commitdiff | tree |
| 2026-07-20 |
Pascal | ui: enable the agentic flow when only the JS sandbox... |
commit | commitdiff | tree |
| 2026-07-20 |
Hongqiang Wang | opencl: Support broadcast for Adreno MUL_MAT and honor... |
commit | commitdiff | tree |
| 2026-07-18 |
Ruixiang Wang | model: rotate injected K/V cache for DFlash (#25823) |
commit | commitdiff | tree |
| 2026-07-18 |
Yash Raj Pandey | llama-quant : exclude i32 ffn_gate_tid2eid routing... |
commit | commitdiff | tree |
| 2026-07-17 |
lhez | opencl: load and use `kernel_gemm_moe_q6_k_f32_ns`... |
commit | commitdiff | tree |
| 2026-07-17 |
Hongqiang Wang | opencl: read/write MoE dp4a activation tiles to local... |
commit | commitdiff | tree |
| 2026-07-17 |
Hongqiang Wang | opencl: transpose q4_K noshuffle scales for coalesced... |
commit | commitdiff | tree |
| 2026-07-17 |
Georgi Gerganov | sync : ggml |
commit | commitdiff | tree |
| 2026-07-17 |
Georgi Gerganov | ggml : bump version to 0.17.0 (ggml/1568) |
commit | commitdiff | tree |
| 2026-07-17 |
fairydreaming | tests : initialize all tensors in test_dsv4_hc to avoid... |
commit | commitdiff | tree |
| 2026-07-17 |
Georgi Gerganov | common : auto-download dflash- and eagle3- HF sidecars... |
commit | commitdiff | tree |
| 2026-07-17 |
Aaron Teo | ggml-blas: default hadamard mul_mat to cpu routine... |
commit | commitdiff | tree |
| 2026-07-17 |
Jeff Bolz | vulkan: Support Q2_0 (#25430) |
commit | commitdiff | tree |
| 2026-07-17 |
Todd Malsbary | sycl: fix row calculation when K_QUANTS_PER_ITERATION... |
commit | commitdiff | tree |
| 2026-07-17 |
Gezahegne | opencl: add ABS op (#25115) |
commit | commitdiff | tree |
| 2026-07-16 |
Hongqiang Wang | opencl: loads quants as uint for q4_K and q5_K flat... |
commit | commitdiff | tree |
| 2026-07-16 |
akleine | docs: added a note about using OpenCl with Adreno 810... |
commit | commitdiff | tree |
| 2026-07-16 |
Aman Gupta | DeepseekV4: Add fused hyper-connection ops (#25585) |
commit | commitdiff | tree |
| 2026-07-16 |
Max Krasnyansky | hexagon: L2 cache handling rework (dirty bit tracking... |
commit | commitdiff | tree |
| 2026-07-16 |
Rajendra Matcha | kleidiai: Add SME vs SME2 distinction in kernel dispatc... |
commit | commitdiff | tree |
| 2026-07-16 |
Ruben Ortlam | vulkan: when using transfer queue for async copies... |
commit | commitdiff | tree |
| 2026-07-16 |
Khashayar Ghafouri | conversion: accept BitNetForCausalLM architecture name... |
commit | commitdiff | tree |
| 2026-07-16 |
Johannes Gäßler | TP: fix Phi3, Bert, Plamo2/3, ChatGLM (#25536) |
commit | commitdiff | tree |
| 2026-07-16 |
Alessandro... | vendor: update BoringSSL to 0.20260713.0 (#25624) |
commit | commitdiff | tree |
| 2026-07-16 |
Aman Gupta | tests: actually exercise `test-recurrent-state-rollback... |
commit | commitdiff | tree |
| 2026-07-16 |
Chipmunk | server : allow text-only slot save/restore with mtmd... |
commit | commitdiff | tree |
| 2026-07-16 |
Ruixiang Wang | convert : fix dflash target tokenizer mismatch during... |
commit | commitdiff | tree |
| 2026-07-16 |
Anav Prasad | CUDA: Support CUDA Virtual Devices (#25228) |
commit | commitdiff | tree |
| 2026-07-16 |
Alexander Heisler | Enable CUDA graphs on volta+turing (#25749) |
commit | commitdiff | tree |
| 2026-07-16 |
Sebastian Dröge | server: Ignore empty / non-existing `Origin` headers... |
commit | commitdiff | tree |
| 2026-07-16 |
liminfei-amd | ggml-cuda : restore prop.integrated on HIP builds ... |
commit | commitdiff | tree |
| 2026-07-16 |
Pranesh Gonegandla | CUDA: dedup MoE gate/up activation quantization (#25441) |
commit | commitdiff | tree |
| 2026-07-16 |
Georgi Gerganov | ci : add official website link to release notes (#25728) |
commit | commitdiff | tree |
| 2026-07-16 |
Georgi Gerganov | quant : allow using manual tensor types with --pure... |
commit | commitdiff | tree |
| 2026-07-16 |
Hongqiang Wang | opencl: disable FA and MoE weights repack to work aroun... |
commit | commitdiff | tree |
| 2026-07-16 |
David Friehs | cuda: extract Q1_0 elements via __byte_perm (#25628) |
commit | commitdiff | tree |
| 2026-07-15 |
Hongqiang Wang | opencl: exclude some moe kernels on Adreno a7x (#25698) |
commit | commitdiff | tree |
| 2026-07-15 |
Aleksander... | ui: Agentic Content UX improvements (#25450) |
commit | commitdiff | tree |
| 2026-07-15 |
fairydreaming | cuda : CUDA GGML_OP_LIGHTNING_INDEXER implementation... |
commit | commitdiff | tree |
| 2026-07-15 |
Adrien Gallouët | tokenize : drop --stdin mutual-exclusion check (#25672) |
commit | commitdiff | tree |
| 2026-07-15 |
Hongqiang Wang | opencl: fix two issues on flash attention for Adreno... |
commit | commitdiff | tree |
| 2026-07-15 |
leonardHONG | CUDA: tighter MMQ src1 buffer size for native fp4 ... |
commit | commitdiff | tree |
| 2026-07-15 |
Gaurav Garg | Fix crash with draft-simple (#25720) |
commit | commitdiff | tree |
| 2026-07-15 |
Pascal | server: fix read_file append_loc space breaking edit_fi... |
commit | commitdiff | tree |
| 2026-07-15 |
Pascal | ui: fix thinking menu never appearing in single-model... |
commit | commitdiff | tree |
| 2026-07-15 |
fairydreaming | cuda : relax tensor contiguity requirements for quantiz... |
commit | commitdiff | tree |
| 2026-07-15 |
Georgi Gerganov | ci : add HF_TOKEN to self-hosted workflows (#25706) |
commit | commitdiff | tree |
| 2026-07-15 |
Pascal | metal: fuse snake activation (mul, sin, sqr, mul, add... |
commit | commitdiff | tree |
| 2026-07-15 |
Michael Lamothe | ggml: add f16 out_prod support for CPU and out_prod... |
commit | commitdiff | tree |
| 2026-07-15 |
Aman Gupta | DeepseekV4: reduce graph splits (#25702) |
commit | commitdiff | tree |
| 2026-07-15 |
Neo Zhang | sycl : fix get_rows Q2_K, Q4_K, Q5_K (#25656) |
commit | commitdiff | tree |
| 2026-07-15 |
Neo Zhang | sycl : support kernel type fp16 for conv2d_dw (#25653) |
commit | commitdiff | tree |
| next |
overview / pkg / ggml / sources / llama.cpp / shortlog