Owner

ggml-org

11 indexed repositories · View on GitHub

llama.cpp
LLM inference in C/C++

C++ · 131903 stars
whisper.cpp
Port of OpenAI's Whisper model in C/C++

C++ · 54535 stars
ggml
Tensor library for machine learning

C++ · 15514 stars
llama.vim
Vim plugin for LLM-assisted code/text completion

Vim Script · 2191 stars
Llama-macOS
A cosy home for your LLMs.

Swift · 1659 stars
llama.vscode
VS Code extension for LLM-assisted code/text completion

TypeScript · 1546 stars
Llama-Windows
Windows Llama App Companion

C# · 254 stars
convert
Convert models to GGUF

Shell · 55 stars
llama-install.sh
Build and install script for llama.app

Python · 35 stars
llama-pages
llama.cpp's official website

Svelte · 21 stars
ggml-org.github.io

Shell · 9 stars

Beginner friendly Everything

Open All

Issues somebody may already be working on are listed last. List everything by date

Misc. bug: docs mention --spec-draft-conf-min, but that flag doesn't exist
Open
- Is something wrong?
bug-unconfirmed

ggml-org/llama.cpp#31118 · Opened Oct 24, 2026

Maintainers usually reply within 1 day
Eval bug: Nemotron Vulkan Intel Core Ultra 5 125H
Open
- Is something wrong?
bug-unconfirmed

ggml-org/llama.cpp#31115 · Opened Oct 24, 2026

Maintainers usually reply within 1 day
Add KailA architecture support
Open
- Is something wrong?
enhancement

ggml-org/llama.cpp#31112 · 1 reaction · Opened Oct 24, 2026

Maintainers usually reply within 1 day
Misc. bug: MUL_MAT: wrong results, exceeds tolerance; gfx1200 HIP/ROCm; f16 + deterministic.
Open
- Is something wrong?
bug-unconfirmed

ggml-org/llama.cpp#31111 · Opened Oct 24, 2026

Maintainers usually reply within 1 day
Misc. bug: SOLVE_TRI, memory fault, gfx1200 HIP/ROCm, intermittent.
Open
- Is something wrong?
bug-unconfirmed

ggml-org/llama.cpp#31108 · Opened Oct 24, 2026

Maintainers usually reply within 1 day
Feature Request: Support for Raon-1 architecture
May be free again A pull request for this issue was closed without being merged. Open
- Is something wrong?
enhancement

ggml-org/llama.cpp#31106 · Opened Oct 24, 2026

Maintainers usually reply within 1 day
Eval bug: ggml-hexagon: vision encoder (mmproj/CLIP) output corrupted depending on image shape on HTP v75 — garbled transcription with repetition loop, correct on CPU
Open
- Is something wrong?
ggml-org/llama.cpp#31104 · 1 comment · Opened Oct 24, 2026

Maintainers usually reply within 1 day
Metal: bf16 MUL_MAT returns wrong results on pre-M5 integrated GPUs (Intel Iris, Common3 family) even with -ngl 0
Open
- Is something wrong?
ggml-org/llama.cpp#31091 · Opened Oct 24, 2026

Maintainers usually reply within 1 day
[Misc] Windows CUDA prebuilt binaries stuck at b11828 while release tags keep advancing (b11846..b11852)
Open
- Is something wrong?
ggml-org/llama.cpp#31085 · Opened Oct 24, 2026

Maintainers usually reply within 1 day
Feature Request: Metis-1.6, Logos & Praxis
Open
- Is something wrong?
enhancement

ggml-org/llama.cpp#31078 · Opened Oct 24, 2026

Maintainers usually reply within 1 day
Eval bug: Gemma 4 31B + draft-mtp: CUDA illegal memory access when MTP drafts right after the KV cache is rolled back or cleared (checkpoint restore or full re-process); no crash with MTP off
Open
- Is something wrong?
bug-unconfirmed

ggml-org/llama.cpp#31077 · 1 comment · Opened Oct 24, 2026

Maintainers usually reply within 1 day
Get ready for qwen4
Open
- Is something wrong?
enhancement

ggml-org/llama.cpp#31072 · 1 comment · 6 reactions · Opened Oct 24, 2026

Maintainers usually reply within 1 day
Feature request: Vulkan shader compression
Open
- Is something wrong?
Difficulty 4/5 3-5 days Newbie friendliness 57/100

ggml-org/ggml#1669 · 2 reactions · Opened Oct 24, 2026

Maintainers usually reply within 1 day
Eval bug: SYCL: XMX-ESIMD path makes MoE prompt processing ~35% slower on Arc Pro B70
Open
- Is something wrong?
ggml-org/llama.cpp#31068 · Opened Oct 24, 2026

Maintainers usually reply within 1 day
Eval bug: SYCL: Segmentation fault on oneAPI 2026.2.0
Open
- Is something wrong?
bug-unconfirmed

ggml-org/llama.cpp#31064 · Opened Oct 23, 2026

Maintainers usually reply within 1 day
gguf: tensor shape overflow check mismatch between gguf_init_from_reader and ggml_new_tensor_impl
Open
- Is something wrong?
Difficulty 2/5 1-3 hours Newbie friendliness 78/100

ggml-org/ggml#1668 · Opened Oct 23, 2026

Maintainers usually reply within 1 day
Eval bug: Running Flint (by Recursive) on GPU (Intel Iris Xe, Windows) using Vulkan, shows ErrorOutOfDeviceMemory even though memory is available
Open
- Is something wrong?
bug-unconfirmed

Difficulty 4/5 3-5 days Newbie friendliness 48/100

ggml-org/llama.cpp#31052 · 1 comment · Opened Oct 23, 2026

Maintainers usually reply within 1 day
Feature Request: Add SOTA on-device TTS model Phonon by Gradium
Open
- Is something wrong?
enhancement

Difficulty 4/5 3-5 days Newbie friendliness 45/100

ggml-org/llama.cpp#31043 · Opened Oct 23, 2026

Maintainers usually reply within 1 day
Misc. bug: Llama delivered via one-line install script provided the Vulkan version instead of the CUDA version
Open
- Is something wrong?
bug-unconfirmed

Difficulty 3/5 1-2 days Newbie friendliness 52/100

ggml-org/llama.cpp#31041 · Opened Oct 23, 2026

Maintainers usually reply within 1 day
[Bug] Bielik-1.5B-v3.0-Instruct Q8_0 generates only `0x1F` bytes on RK3588 (aarch64) — Q4_K_M works, other Q8_0 models work
Open
- Is something wrong?
Difficulty 4/5 3-5 days Newbie friendliness 48/100

ggml-org/llama.cpp#31039 · Opened Oct 23, 2026

Maintainers usually reply within 1 day
Compile bug: SYCL AOT build for mtl-m (Xe-LPG) fails: `dpas.nosrc0` intrinsic not supported, introduced by d81d0d8fb
Open
- Is something wrong?
bug-unconfirmed

Difficulty 4/5 1-2 days Newbie friendliness 48/100

ggml-org/llama.cpp#31033 · Opened Oct 23, 2026

Maintainers usually reply within 1 day
Misc. bug: SYCL bf conversion instruction not supported! Arrow Lake iGPU
Open
- Is something wrong?
bug-unconfirmed

Difficulty 4/5 1-2 days Newbie friendliness 49/100

ggml-org/llama.cpp#31029 · 1 comment · Opened Oct 23, 2026

Maintainers usually reply within 1 day
Eval bug: Significant performance difference between llama-prism and main llama.cpp builds (same base version)
Open
- Is something wrong?
bug-unconfirmed

Difficulty 4/5 3-5 days Newbie friendliness 38/100

ggml-org/llama.cpp#31016 · 5 comments · Opened Oct 23, 2026

Maintainers usually reply within 1 day
whisper-cli: output to stdout on windows
Open
- Is something wrong?
Difficulty 1/5 Under an hour Newbie friendliness 80/100

ggml-org/whisper.cpp#4126 · Opened Oct 23, 2026

Maintainers usually reply within 4 days
Eval bug: Flint NVFP4: tool-call XML remains in reasoning, turn ends with stop (llama.cpp b11823 / OpenCode 2.1.2)
Open
- Is something wrong?
Difficulty 5/5 Over a week Newbie friendliness 32/100

ggml-org/llama.cpp#31000 · 1 comment · Opened Oct 22, 2026

Maintainers usually reply within 1 day
Misc. bug: https://llama.app/install.sh doesn't add `~/.local/bin` to path, cannot find llama after install
Open
- Is something wrong?
bug-unconfirmed

Difficulty 2/5 1-3 hours Newbie friendliness 64/100

ggml-org/llama.cpp#30998 · Opened Oct 22, 2026

Maintainers usually reply within 1 day
Compile bug: wavefront shifts are not supported on GFX10+
Open
- Is something wrong?
bug-unconfirmed

Difficulty 3/5 1-2 days Newbie friendliness 57/100

ggml-org/llama.cpp#30989 · Opened Oct 22, 2026

Maintainers usually reply within 1 day
OpenCL (Adreno): kernels fail to build on Adreno X1-45 (Snapdragon X X1-26-100, Windows) — driver lacks cl_khr_subgroups / cl_qcom_subgroup_shuffle
Open
- Is something wrong?
bug-unconfirmed

Difficulty 4/5 3-5 days Newbie friendliness 52/100

ggml-org/llama.cpp#30983 · Opened Oct 22, 2026

Maintainers usually reply within 1 day
Misc. bug: llama-server teardown hangs after 'cleaning up before exit...' on first SIGTERM (spec decoder + quantized draft cache, CUDA); second signal terminates
Open
- Is something wrong?
Difficulty 4/5 3-5 days Newbie friendliness 43/100

ggml-org/llama.cpp#30975 · Opened Oct 22, 2026

Maintainers usually reply within 1 day
llama serve exits with "cleaning up before exit" ~30-60s after startup when launched from Finder
Open
- Is something wrong?
Difficulty 2/5 1-3 hours Newbie friendliness 76/100

ggml-org/Llama-macOS#131 · Opened Oct 22, 2026
Copyright of examples/wchess (Debian packaging)
Open
- Is something wrong?
Difficulty 1/5 Under an hour Newbie friendliness 35/100

ggml-org/whisper.cpp#4125 · Opened Oct 22, 2026

Maintainers usually reply within 4 days
Eval bug: RPC + Vulkan tensor splitting crash: reproducible lifetime bug and one-line fix
Open
- Is something wrong?
bug-unconfirmed

Difficulty 4/5 3-5 days Newbie friendliness 50/100

ggml-org/llama.cpp#30954 · 1 comment · Opened Oct 22, 2026

Maintainers usually reply within 1 day
GGML MKL backend silently falls back to single-threaded execution inside OpenMP parallel regions
Open
- Is something wrong?
Difficulty 4/5 3-5 days Newbie friendliness 48/100

ggml-org/ggml#1667 · Opened Oct 22, 2026

Maintainers usually reply within 1 day
Eval bug: CUDA gpt-oss-120b with IQ1_M (or TQ1_0/TQ2_0) experts produces NaN ("GGGG") at prompt batch > 8: fp16 accumulation and result in the dequantize + cuBLAS fallback overflow
Open
- Is something wrong?
Difficulty 4/5 3-5 days Newbie friendliness 42/100

ggml-org/llama.cpp#30951 · 1 comment · Opened Oct 21, 2026

Maintainers usually reply within 1 day
Eval bug: CUDA Q4_1/Q5_1 mat-vec kernels overflow fp16 when multiplying the block offset by the Q8_1 block sum (gpt-oss emits "G", NaN logits)
May be free again A pull request for this issue was closed without being merged. Open
- Is something wrong?
Difficulty 3/5 1-2 days Newbie friendliness 55/100

ggml-org/llama.cpp#30950 · 2 comments · Opened Oct 21, 2026

Maintainers usually reply within 1 day
App's own requests fail when --api-key is set via extraServerArgs
Open
- Is something wrong?
Difficulty 4/5 3-5 days Newbie friendliness 54/100

ggml-org/Llama-macOS#129 · Opened Oct 21, 2026
DeepSeek-V4.1-Flash (PR #30817): single-token output corruption when a cached prompt prefix is reused (Engram n-gram history)
Open
- Is something wrong?
Difficulty 4/5 3-5 days Newbie friendliness 38/100

ggml-org/llama.cpp#30944 · Opened Oct 21, 2026

Maintainers usually reply within 1 day
prefill: ~7% pp512 regression on Arc B580 since #30160 (embeddings get_rows reorder), constant per-invocation cost
Open
- Is something wrong?
Difficulty 4/5 3-5 days Newbie friendliness 48/100

ggml-org/llama.cpp#30939 · Opened Oct 21, 2026

Maintainers usually reply within 1 day
Eval bug: terminate called after throwing an instance of 'common_json_error'
Open
- Is something wrong?
bug-unconfirmed

Difficulty 3/5 1-2 days Newbie friendliness 48/100

ggml-org/llama.cpp#30938 · Opened Oct 21, 2026

Maintainers usually reply within 1 day
Misc. bug: "CommandLine Error: Option 'sanitizer-early-opt-ep' registered more than once!"
Open
- Is something wrong?
bug-unconfirmed

Difficulty 4/5 3-5 days Newbie friendliness 48/100

ggml-org/llama.cpp#30937 · Opened Oct 21, 2026

Maintainers usually reply within 1 day
Feature Request: EAGLE-4 support
Open
- Is something wrong?
enhancement

Difficulty 5/5 Over a week Newbie friendliness 25/100

ggml-org/llama.cpp#30925 · Opened Oct 21, 2026

Maintainers usually reply within 1 day
Vulkan: Mali-G710 (Android 16, VK 1.1) still crashes in QueueSubmit after applying #28593 null-pointer fixes
Open
- Is something wrong?
Difficulty 4/5 1-2 days Newbie friendliness 42/100

ggml-org/llama.cpp#30920 · Opened Oct 21, 2026

Maintainers usually reply within 1 day
CPU backend: garbage output (repetition loops) for the qwen35 hybrid family (Qwen3.8-27B / Holo4), while the Vulkan backend serves the same binary and GGUF correctly
Open
- Is something wrong?
Difficulty 4/5 3-5 days Newbie friendliness 45/100

ggml-org/llama.cpp#30906 · Opened Oct 21, 2026

Maintainers usually reply within 1 day
Eval bug: Laguna repeats <assistant> on Metal for ubatches >= 32 tokens (ffn_moe_down input exceeds F16 range)
Open
- Is something wrong?
Difficulty 2/5 1-3 hours Newbie friendliness 78/100

ggml-org/llama.cpp#30896 · Opened Oct 20, 2026

Maintainers usually reply within 1 day
Alias does not work
Open
- Is something wrong?
Difficulty 3/5 1-2 days Newbie friendliness 55/100

ggml-org/Llama-macOS#128 · Opened Oct 20, 2026
Eval bug: SYCL F16: NaN output on dense Qwen3.8-27B Q4_K_M since #29107 (IQ3 code reorder); fused GEMM off fixes it (Arc Pro B70)
Open
- Is something wrong?
Difficulty 4/5 3-5 days Newbie friendliness 48/100

ggml-org/llama.cpp#30892 · 6 comments · Opened Oct 20, 2026

Maintainers usually reply within 1 day
Vulkan (AMD, Windows, no ReBAR): DEVICE_LOCAL|HOST_VISIBLE buffers are backed by system memory and exhaust the WDDM shared budget -> vkQueueSubmit returns ErrorOutOfDeviceMemory
Open
- Is something wrong?
Difficulty 4/5 3-5 days Newbie friendliness 54/100

ggml-org/llama.cpp#30888 · 1 comment · Opened Oct 20, 2026

Maintainers usually reply within 1 day
Qwen3.8-Flash-Next + DFlash2 crashes on first prompt in extract_layer_inputs()
Open
- Is something wrong?
Difficulty 4/5 3-5 days Newbie friendliness 42/100

ggml-org/llama.cpp#30887 · 1 comment · Opened Oct 20, 2026

Maintainers usually reply within 1 day
Flash attention (CUDA/HIP): garbage output at long context on gfx1201 (RDNA4) - immediate EOS, head_dim 256
Open
- Is something wrong?
Difficulty 4/5 3-5 days Newbie friendliness 35/100

ggml-org/llama.cpp#30884 · 3 comments · Opened Oct 20, 2026

Maintainers usually reply within 1 day
Misc. bug: Sarvam 30B MoE converter labels byte-fallback tokens <0x00>–<0xFF> NORMAL instead of BYTE
Open
- Is something wrong?
bug-unconfirmed

Difficulty 2/5 1-3 hours Newbie friendliness 77/100

ggml-org/llama.cpp#30879 · Opened Oct 20, 2026

Maintainers usually reply within 1 day

Showing the newest 100

This page lists what was indexed most recently. Advanced filter has the whole inventory, narrowed by language, difficulty and how long a task takes.

Open advanced filter

ggml-org

Is something wrong?

Is something wrong?

Is something wrong?

Is something wrong?

Is something wrong?

Is something wrong?

Is something wrong?

Is something wrong?

Is something wrong?

Is something wrong?

Is something wrong?

Is something wrong?

Is something wrong?

Is something wrong?

Is something wrong?

Is something wrong?

Is something wrong?

Is something wrong?

Is something wrong?

Is something wrong?

Is something wrong?

Is something wrong?

Is something wrong?

Is something wrong?

Is something wrong?

Is something wrong?

Is something wrong?

Is something wrong?

Is something wrong?

Is something wrong?

Is something wrong?

Is something wrong?

Is something wrong?

Is something wrong?

Is something wrong?

Is something wrong?

Is something wrong?

Is something wrong?

Is something wrong?

Is something wrong?

Is something wrong?

Is something wrong?

Is something wrong?

Is something wrong?

Is something wrong?

Is something wrong?

Is something wrong?

Is something wrong?

Is something wrong?

Is something wrong?

Showing the newest 100

Get new issues in your inbox