Search results

GGML_NATIVE appends -march=native after user GGML_ARCH_FLAGS, breaking fullfp16 (iqk) builds on Apple Silicon

Status: Open.

#2135 In ikawrakow/ik_llama.cpp;

· hchengit opened on Jul 15, 2026

Feature Request: ik_llama docker 'swap' version seems to use an old UI?

enhancementNew feature or requestNew feature or request

Status: Open.

#2122 In ikawrakow/ik_llama.cpp;

· StudioNirin opened on Jul 13, 2026

Feature Request: Support Step 3.7 Flash MTP

enhancementNew feature or requestNew feature or request

Status: Open.

#2091 In ikawrakow/ik_llama.cpp;

· everson opened on Jul 6, 2026

Bug: llama-server hangs when used with Hermes: CUDA keeps occupied, no output and frozen inference logs

Status: Open.

#2088 In ikawrakow/ik_llama.cpp;

· zhai-hello opened on Jul 6, 2026

DFlash: improve SWA implementation

enhancementNew feature or requestNew feature or request

Status: Open.

#2052 In ikawrakow/ik_llama.cpp;

· ikawrakow opened on Jun 29, 2026

Bug: KV cache prompt comparison runs before checkpoint restore, causing erroneous re-evaluation (with proposed fix)

Status: Open.

#2033 In ikawrakow/ik_llama.cpp;

· syncword opened on Jun 25, 2026

Bug: GGML_SCHED_MAX_COPIES=2 produces repeated tokens when graph reuse is enabled

Status: Open.

#2006 In ikawrakow/ik_llama.cpp;

· KeinNiemand opened on Jun 20, 2026

Feature Request: Add support for Gemma4-12b unified vision and audio projectors

enhancementNew feature or requestNew feature or request

Status: Open.

#2002 In ikawrakow/ik_llama.cpp;

· aadametz opened on Jun 19, 2026

Bug: CUDA illegal memory access in ggml_backend_cuda_synchronize

Status: Open.

#1975 In ikawrakow/ik_llama.cpp;

· glibg10b opened on Jun 15, 2026

Hybrid (qwen3next / qwen35moe) recurrent-state corruption under concurrent decoding (np>=3): root cause + fix

Status: Open.

#1932 In ikawrakow/ik_llama.cpp;

· poisonxa16 opened on Jun 7, 2026

Bug: Maybe MTP cache invalidation?

Status: Open.

#1931 In ikawrakow/ik_llama.cpp;

· rhossack opened on Jun 7, 2026

Bug: newest nVidia driver 610.47 fails to run llama-server with "munmap_chunk(): invalid pointer" error

Status: Open.

#1909 In ikawrakow/ik_llama.cpp;

· Iipal opened on Jun 1, 2026