- v0.7.0 (2026-10-15)
Overview llama.cpp v0.7.0 adds the LiliCorr speculative decoding technique and MTP for GLM5-Next, a new GPU cache for MoE experts kept in host memory ( --moe-cache-mib , up to ~2.2x decode speedup when experts are offloaded), and a large batch of new models: K2 Horizon (dense and
- b11669 (2026-10-15)
llama.cpp : bump version to 0.7.0 ( #30603 ) Website: https://llama.app Attestations: https://github.com/ggml-org/llama.cpp/attestations/56338993 macOS/iOS: macOS Apple Silicon (arm64) macOS Apple Silicon (arm64, KleidiAI enabled) DISABLED macOS Intel (x64) iOS XCFramework Linux:
- b11662 (2026-10-15)
sync : ggml Website: https://llama.app Attestations: https://github.com/ggml-org/llama.cpp/attestations/56327800 macOS/iOS: macOS Apple Silicon (arm64) macOS Apple Silicon (arm64, KleidiAI enabled) DISABLED macOS Intel (x64) iOS XCFramework Linux: Ubuntu x64 (CPU) Ubuntu arm64 (C
- b11658 (2026-10-15)
models : add maion-coder ( #29778 ) Add Support to maion-coder arch iAdd Support to support Draft models Address comments Remove maion context shift function Co-authored-by: Karthik Vijayan
[email protected] Website: https://llama.app Attestations: https://github.com/ggml-org
- b11655 (2026-10-15)
vulkan : fix concurrent first device init ( #29766 ) vulkan : fix concurrent first device init tests : load models in parallel in test-thread-safety Loading them serially never exercised concurrent first backend init, so the crash reported in #29764 was invisible to this test. vu
- b11647 (2026-10-15)
vulkan : copy strided f16 KV once on AMD so reads spread across memory channels ( #27703 ) vulkan : copy strided f16 KV once so reads spread across memory channels Assisted-by: Claude (Opus 4.8 / Opus 5) tests : cover the f16 KV copy to scratch Assisted-by: Claude (Opus 4.8 / Opu
- b11645 (2026-10-15)
metal : add BF16 variant of the tensor API flash attention kernel ( #30491 ) metal : add BF16 variant of the tensor API flash attention kernel cont : move tensor FA kernels to kernels/fa_tensor.metal Website: https://llama.app Attestations: https://github.com/ggml-org/llama.cpp/a
- b11643 (2026-10-15)
graph: keep the graph non empty without embeddings ( #30522 ) An embedding-only graph has no logits, and build_pooling() returns early when embeddings are off, so since the models no longer expand t_embd the graph ends up empty and its inputs are never allocated. Expand t_embd in
- b11642 (2026-10-15)
server: save only restorable slot states to the prompt cache ( #30498 ) server: save only restorable slot states to the prompt cache The idle-slot save loop stores every idle slot whenever a task starts, while only completion tasks restore from the prompt cache. Rerank, embedding
- b11644: kv-cache: honor PARTIAL_ONLY in the plain KV cache (#30514) (2026-10-15)
llama: honor PARTIAL_ONLY in the plain KV cache A non-SWA llama_kv_cache rolls back with seq_rm, so it holds no partial state, yet state_write ignored the flag and serialized every cell. A draft context made of a plain KV cache (MTP head, dense -md draft) therefore copied its who