This repository is a fork of llama.cpp with better CPU and hybrid GPU/CPU performance, new SOTA quantization types, first-class Bitnet support, better DeepSeek performance via MLA, FlashMLA, fused MoE operations and tensor overrides for hybrid GPU/CPU inference, row-interleaved quant packing, etc.

Latest News

Model Support

LlaMA-3-Nemotron PR 377, Qwen3 PR 355, GLM-4 PR 344, Command-A PR 341, bitnet-b1.58-2B-4T PR 337, LLaMA-4 PR 321], Gemma3 PR 276, DeepSeek-V3 PR 176

Quantization improvements

IQ1_M PR 327, IQ2_XS PR 312, Q2_K, Q4_K, Q5_K, Q4_1, Q5_1 PR 302, Q4_0, Q5_0, Q6_0, Q3_K, Q6_K, IQ4_XS, IQ4_NL PR 295,

Features

May 12 2025: User can now control if/which operations with tensors held in RAM are offloaded to the GPU. See PR 405
May 12 2025: Compatibility issues with mainline llama.cpp GGUFs for DeepSeek models with MLA enabled were resolved in PR 394. The lower prompt processing performance resulting from using llama.cpp-style MLA GGUFs was recovered in PR 409.
April 21 2025: ik_llama.cpp builds and runs successfully on Android (using termux), see PR 336
March 1 2025: Smart Expert Reduction for faster DeepSeek inference, see
Feb 25 2025: Tensor overrides for better control where model weights are stored (GPU or CPU)
Feb 23 2025: sweep-bench - better performance benchmarking
Feb 19 2025: Q8_KV - new type for 8-bit KV-cache quantization
March 7 2025: Custom quantization mixes using regular expressions
Feb 25 2025: Tensor overrides for better control where model weights are stored (GPU or CPU)

Performance improvements

May 11 2025: 🚀 Slightly faster flash attention for DeepSeek models on CUDA, along with extending compatibility to Touring or newer GPUs. See PR 408
May 7 2025: 🚀 Faster TG for DeepSeek models with GPU or hybrid GPU/CPU inference. See PR 386 for details. Caveat: Ampere or newer Nvidia GPU required
May 4 2025: 🚀 Significant token generation performance improvement on CUDA with Flash Attention for GQA models. For details and benchmarks see PR #370
April 17 2025: 🚀 Better CPU Flash Attention token generation performance, see PR 332
April 3 2025: 🚀 Much faster MoE implementation on Metal, see PR 307
March 25 2025: 🚀 Better MoE performance on CUDA
March 23 2025: 🚀 Better batched processing speed for DeepSeek models
March 18 2025: Reduce compute buffer size
March 10 2025: 🚀 Better TG performance for MoE models on CUDA
Feb 23 2025: 🚀 Fused FFN ops for faster MoE inference
Feb 20 2025: 🚀 Fast GEMM/GEMV for IQ1_S

Flash-MLA

March 21 2025: 🚀 FlashMLA-3: fastest CPU-only inference for DeepSeek models see
March 17 2025: 🚀 FlashMLA-2 performance improvements
March 12 2025: Allow Q8_0 KV cache with FlashMLA-2 on CUDA
March 9 2025: 🚀 FlashMLA on CUDA
March 8 2025: 🚀 Faster FlashMLA CPU implementation
March 3 2025: 🚀 Introducing FlashMLA - MLA with Flash Attention
Feb 27 2025: MLA without transposed cache
Feb 13 2025: Allow Q8_0 quantized cache with MLA
Feb 11 2025: 🚀 Flash Attention support for DeepSeek models
Feb 9 2025: 🚀 MLA for DeepSeek models

Resources

There is no single point of reference describing all new ik_llama.cpp features. Pull requests often contain detailed information, so browsing the PRs is often the best way to learn about new features and how to use them. In addition

The Wiki page has performance comparisons to mainline llama.cpp
This guide is a good place to start if you came here because of DeepSeek models
This discussion is about running DeepSeek-V3/R1 on a 16 x 3090 setup
This discussion describes the new quantization types available in ik_llama.cpp

Contributing

Contributions in form of pull requests, issue submissions (bug reports, feature requests), or general discussions, are welcome.

License

MIT

Languages

C++ 55.5%

C 16.2%

Cuda 14.4%

Python 5.4%

Metal 3%

Other 5.5%