Releases·Tool Release
llama.cpp v0.5.0 adds HRM-Text/MiMo-V2.6 support, CUDA and Metal fusion optimizations, ggml 0.25.0
ggml-org released llama.cpp v0.5.0 on 2026-09-23, focusing on backend performance and correctness, broader model coverage, and more robust server/router operation. The release bundles ggml 0.25.0, CUDA and Metal/SYCL/Vulkan optimizations, new model support (HRM-Text/DFM Mimir 1B, MiMo-V2.6, HunyuanOCR DFlash), and API changes like LoRA loading from an open FILE handle. Practitioners running local inference gain faster attention and MoE paths plus more stable multi-model serving.
1 engagement1 sourceWed, Sep 23, 2026, 08:50 PM
AI Overview
- llama.cpp v0.5.0 (2026-09-23) adds HRM-Text/DFM Mimir 1B support plus MiMo-V2.6 and HunyuanOCR DFlash conversion support
- The release accelerates CUDA conv2d with implicit GEMM and adds Metal MoE and SSM_CONV fusion optimizations (#29135, #28948)
- Server changes: --host accepts comma-separated TCP addresses and UNIX sockets, function-call outputs gain input_image support
- ggml 0.25.0 expands hyper-connection, flash-attention, and fused MoE/SSM backend support, with RPC protocol v7
- New model handling extends Nemotron MTP/Nemotron-H, adds Qwen4Exp hyper-connection ops and sparse flash attention
- Fixes cover router eviction races, SigLIP buffer overruns on tall/wide images, and chat parsers for Ling 3.0, DeepSeek V3.2/V4, and Gemma 4
Related
Releases·Tool Release
llama.cpp v0.4.0 adds Qwen3.8-Flash-Next and Nemotron-3-Puzzle-75B-A9B support, lazy tensor loading, ggml 0.23.0
llama.cpp shipped v0.4.0, adding initial support for Qwen3.8-Flash-Next and Nemotron-3-Puzzle models. The release also brings on-demand tensor reading, per-slot server context limits, video input options, and a ggml update to 0.23.0 featuring major sparse flash attention and RDMA work.
llama.cpp shipped v0.4.0, adding initial support for Qwen3.8-Flash-Next and Nemotron-3-Puzzle models. The release also brings on-demand tensor reading, per-slot server context limits, video input options, and a ggml update to 0.23.0 featuring major sparse flash attention and RDMA work.
6 engagement1 sourcegithubFri, Sep 4, 2026, 07:56 PM
Releases·Tool Release
llama.cpp v0.7.0 adds LiliCorr speculative decoding, MoE host-memory GPU cache, and GLM5-Next MTP
ggml-org shipped llama.cpp v0.7.0 on 2026-10-15. The release adds LiliCorr speculative decoding, MTP for GLM5-Next, and a host-memory GPU cache for MoE experts offering up to ~2.2x decode speedup. It also expands model support and revamps the server and web UI, making it a notable upgrade for local inference practitioners.
ggml-org shipped llama.cpp v0.7.0 on 2026-10-15. The release adds LiliCorr speculative decoding, MTP for GLM5-Next, and a host-memory GPU cache for MoE experts offering up to ~2.2x decode speedup. It also expands model support and revamps the server and web UI, making it a notable upgrade for local inference practitioners.
0 engagement1 sourcegithubThu, Oct 15, 2026, 05:13 PM
Releases·Tool Release
llama.cpp releases v0.2.0 with nightly build b10566
llama.cpp v0.2.0 is out, featuring a new nightly build and a release script for preparation. The update syncs with ggml v0.21.0 and includes CI improvements.
llama.cpp v0.2.0 is out, featuring a new nightly build and a release script for preparation. The update syncs with ggml v0.21.0 and includes CI improvements.
142 engagement2 sourcesgithub, redditFri, Aug 21, 2026, 06:32 PM
Releases·Tool Release
llama.cpp v0.4.1 adds Maple 20B-A1B, Tencent Hy 4, Spark2.5 support and bumps ggml to v0.24.0
The llama.cpp project shipped v0.4.1 on GitHub, adding support for Maple 20B-A1B (ternary MoE, CPU), Tencent Hy 4, and Spark2.5, plus Kimi-K3 recurrent-state rollback. The release refactors JSON schema handling and chat parsing, fixes MTP KV cache and GDN normalization bugs, and bumps ggml to v0.24.0 with expanded backend coverage. It matters to anyone running local inference on the affected model families.
The llama.cpp project shipped v0.4.1 on GitHub, adding support for Maple 20B-A1B (ternary MoE, CPU), Tencent Hy 4, and Spark2.5, plus Kimi-K3 recurrent-state rollback. The release refactors JSON schema handling and chat parsing, fixes MTP KV cache and GDN normalization bugs, and bumps ggml to v0.24.0 with expanded backend coverage. It matters to anyone running local inference on the affected model families.
15 engagement1 sourcegithubMon, Sep 14, 2026, 06:27 PM
Releases·Tool Release
llama.cpp dev lands three Metal MoE PRs for decode, prefill, and KV caching, tested on Qwen3-30B-A3B
A LocalLLaMA developer with an M5 MacBook Pro 24GB opened three llama.cpp pull requests improving Metal MoE decode, prefill, and caching, tested on Qwen3-30B-A3B. Because testing is limited to an M5 and an older M1 Pro 32GB, the author is recruiting community testers, including for Qwen3.8 Flash Next.
A LocalLLaMA developer with an M5 MacBook Pro 24GB opened three llama.cpp pull requests improving Metal MoE decode, prefill, and caching, tested on Qwen3-30B-A3B. Because testing is limited to an M5 and an older M1 Pro 32GB, the author is recruiting community testers, including for Qwen3.8 Flash Next.
1 engagement1 sourceredditThu, Sep 3, 2026, 02:13 PM