TileLang

llama.cpp is a C/C++ implementation for running large language and vision-language model inference locally or in the cloud, including on CPUs, GPUs, laptops, Apple silicon devices, and small computers such as Raspberry Pis. The project is built on the ggml library and uses integer quantization from 1.5-bit through 8-bit to reduce memory use and accelerate inference, with CPU, CUDA, HIP, Metal, Vulkan, SYCL, OpenCL, and other hardware backends. It supports hybrid CPU-GPU inference for models larger than available VRAM and includes command-line tools, a built-in web interface, and an OpenAI-compatible REST API server. The project is distributed as source code, Docker images, and pre-built binaries, with installation instructions at llama.app.

XLA (Accelerated Linear Algebra) is an open-source machine-learning compiler for GPUs, CPUs, and other ML accelerators. It takes models from frameworks such as JAX, PyTorch, and TensorFlow, traces and compiles them into optimized executable code for the target hardware. The project is developed under OpenXLA and is primarily used through the documentation and integration of the corresponding ML framework; its repository is intended mainly for compiler contributors and frontend or hardware-backend integrators.

Project Titania is a from-scratch large language model system covering a decoder-only transformer, compiler, instruction-set architecture, ISA simulator, and planned GPU hardware. It runs Qwen3-0.6B and expresses the model as GPU kernels for matrix multiplication, normalization, attention, and activation functions. The compiler lowers those kernels into Titania ISA programs, while the ISA simulator executes the instructions as the reference implementation of their behavior. Running `titania run --device isasim` compiles each model operation into a Titania kernel for its shape and displays the simulated kernel execution, including a warp's current disassembly. The repository also provides a CLI that can run the model on the CPU or simulator and includes a bash tool whose commands run without confirmation. The project is written to keep each layer readable and uses the MIT license. Its hardware layer is an RTL GPU design planned to progress through RTL simulation, FPGA synthesis, and eventual silicon; the current end-to-end model execution described by the repository uses the CPU and ISA simulator.

KTransformers is an open-source research framework for efficient large-language-model inference and fine-tuning through heterogeneous CPU-GPU computing. Its user-facing capabilities are provided through the kt-kernel source tree: high-performance inference and supervised fine-tuning (SFT). For mixture-of-experts models, it uses heterogeneous expert scheduling, placing frequently used or “hot” experts on GPUs and other experts on CPUs with optimized AMX, AVX512, or AVX2 kernels and NUMA-aware memory management. It supports CPU-side INT4/INT8 quantized weights, GPU-side GPTQ, native BF16 and FP8 precision, and multiple CPU, GPU, and NPU backends. The SFT capability integrates with LlamaFactory for hybrid CPU-GPU fine-tuning of large MoE models, including LoRA and full-parameter training, with INT4/INT8 quantization options.

vLLM is an open-source library and inference-serving engine for deploying large language models on GPUs and other hardware accelerators. Originally developed in UC Berkeley's Sky Computing Lab, it is maintained by a broad community and provides an OpenAI-compatible API server, as well as Anthropic Messages API and gRPC support. Its serving stack manages attention key-value memory with PagedAttention, continuous batching, chunked prefill, prefix caching, streaming generation, and disaggregated prefill, decode, and encode. It also supports speculative decoding, quantization, optimized attention and GEMM/MoE kernels, automatic kernel generation and graph-level transformations, structured outputs, tool calling, multiple decoding algorithms, and distributed inference through tensor, pipeline, data, expert, and context parallelism. The project integrates with Hugging Face models and supports decoder-only, mixture-of-experts, hybrid state-space, multimodal, embedding, retrieval, reward, and classification models. It can run across NVIDIA, AMD, Intel, and other supported accelerators, as well as x86, ARM, and PowerPC CPUs, and can be installed with uv or pip or built from source.