Search Results for "inference engine"
Sort By:
Showing 106 open source projects for "inference engine"
View related business solutions-
$300 Free Credits to Build on Google CloudPut your $300 in credit toward real workloads, then keep building with free monthly usage for 20+ products. No commitment and no charge until you upgrade.
-
Custom VMs From 1 to 96 vCPUs With 99.95% UptimeLive migration and automatic failover keep workloads online through maintenance. One free e2-micro VM every month.
-
1
Transformer Engine
A library for accelerating Transformer models on NVIDIA GPUs
Transformer Engine (TE) is a library for accelerating Transformer models on NVIDIA GPUs, including using 8-bit floating point (FP8) precision on Hopper GPUs, to provide better performance with lower memory utilization in both training and inference. TE provides a collection of highly optimized building blocks for popular Transformer architectures and an automatic mixed precision-like API that can be used seamlessly with your framework-specific code. -
2
Temporal Inference Engine
A real time inference engine for temporal logical specifications
A real time inference engine for temporal logical specifications, which is able to acquire, process and generate any binary or real signal through POSIX IPC, files or UNIX sockets. Specifications of signals and dynamic systems are represented as special graphs and executed in real time, with a predictable sampling time of few milliseconds. Real time signal processing, dynamic system control, state machine modeling and logical property verification are some fields of application of this software. ... -
3
MLX Engine
LM Studio Apple MLX engine
MLX Engine is the Apple MLX-based inference backend used by LM Studio to run large language models efficiently on Apple Silicon hardware. Built on top of the mlx-lm and mlx-vlm ecosystems, the engine provides a unified architecture capable of supporting both text-only and multimodal models. Its design focuses on high-performance on-device inference, leveraging Apple’s MLX stack to accelerate computation on M-series chips. -
4
ReactiveMP.jl
High-performance reactive message-passing based Bayesian engine
ReactiveMP.jl is a Julia package that provides an efficient reactive message passing based Bayesian inference engine on a factor graph. The package is a part of the bigger and user-friendly ecosystem for automatic Bayesian inference called RxInfer. While ReactiveMP.jl exports only the inference engine, RxInfer provides convenient tools for model and inference constraints specification as well as routines for running efficient inference both for static and real-time datasets. -
Ship Agents FasterGemini Enterprise Agent Platform lets you rapidly build, scale, govern and optimize production-ready agents grounded in your organization's data. The platform enables developers to build custom or pre-built agents for virtually any use case. New customers get $300 in free credits.
-
5
ds4.c
DeepSeek 4 Flash local inference engine for Metal
ds4.c is a specialized local inference engine created by antirez for running DeepSeek V4 Flash models directly on Apple Silicon hardware using Metal acceleration. Unlike general-purpose inference runtimes, the project is intentionally optimized for a specific model family, enabling highly efficient execution and simplified architecture. The engine includes DS4-specific model loading, KV cache management, prompt rendering, and OpenAI-compatible server APIs for local deployment workflows. ... -
6
uzu
A high-performance inference engine for AI models
uzu is a high-performance inference engine designed to run artificial intelligence models efficiently on Apple Silicon hardware. Written primarily in Rust and leveraging Apple’s Metal framework, the project focuses on maximizing performance when executing large language models and other AI workloads on devices such as Mac computers with M-series chips. The engine implements a hybrid architecture in which model layers can be executed either as custom GPU kernels or through Apple’s MPSGraph API, allowing it to balance performance and compatibility depending on the workload. ... -
7
vLLM
A high-throughput and memory-efficient inference and serving engine
vLLM is a fast and easy-to-use library for LLM inference and serving. High-throughput serving with various decoding algorithms, including parallel sampling, beam search, and more. -
8
Nano-vLLM
A lightweight vLLM implementation built from scratch
Nano-vLLM is a lightweight implementation of the vLLM inference engine designed to run large language models efficiently while maintaining a minimal and readable codebase. The project recreates the core functionality of vLLM in a simplified architecture written in approximately a thousand lines of Python, making it easier for developers and researchers to understand how modern LLM inference systems work. -
9
SimpleLLM
950 line, minimal, extensible LLM inference engine built from scratch
SimpleLLM is a minimal, extensible large language model inference engine implemented in roughly 950 lines of code, built from scratch to serve both as a learning tool and a research platform for novel inference techniques. It provides the core components of an LLM runtime—such as tokenization, batching, and asynchronous execution—without the abstraction overhead of more complex engines, making it easier for developers and researchers to understand and modify. ... -
Build Securely on AWS with Proven FrameworksMoving to the cloud brings new challenges. How can you manage a larger attack surface while ensuring great network performance? Turn to Fortinet’s Tested Reference Architectures, blueprints for designing and securing cloud environments built by cybersecurity experts. Learn more and explore use cases in this white paper.
-
10
Kimi K3 in C
A 2.78-trillion-parameter Kimi K3 running inference on a single CPU
Kimi K3 in C is a portable C99 inference engine built to run the 2.78-trillion-parameter Kimi K3 model on CPUs without BLAS, machine-learning frameworks, or GPUs. It demonstrates inference from a roughly 1.56 TB checkpoint with measured memory use as low as 8.24 GB. The runtime streams model trunk layers and routed experts from disk instead of keeping all weights resident in memory. -
11
llama.cpp
LLM inference in C/C++
llama.cpp is a high-performance C and C++ project for running large language models locally and in the cloud with minimal setup. It is built around efficient inference, broad hardware support, and the GGUF model format. The project supports many model families and has become a major foundation for local AI tools, model serving, and embedded inference workflows. It provides command-line tools, a server mode with an OpenAI-compatible API style, model conversion utilities, and extensive backend... -
12
Jlama
Jlama is a modern LLM inference engine for Java
Jlama is a modern inference engine written entirely in Java that enables developers to run large language models locally within Java applications. Unlike frameworks that require external APIs or remote services, Jlama performs inference directly on a machine using pre-trained models. This allows organizations to integrate generative AI features into their systems while maintaining full control over data privacy and infrastructure. -
13
BaseRT
Fastest LLM inference runtime for Apple Silicon
...Its server implements OpenAI-compatible chat, completion, embedding, transcription, tool-calling, and multimodal endpoints. The custom .base format supports affine quantization from Q2 through Q8, optional AWQ calibration, and signed model bundles. Stable C interfaces connect the engine with Python, Node.js, Rust, and Swift applications. The repository contains the open CLI, format specifications, bindings, documentation, and benchmarks, while the prebuilt inference engine uses a separate license. -
14
TokenSpeed
TokenSpeed is a speed-of-light LLM inference engine
TokenSpeed is an LLM inference engine designed for high-performance production agent workloads. It aims to combine TensorRT-LLM-level speed with vLLM-level usability, making it relevant for teams that need fast generation without sacrificing developer ergonomics. The project is focused on the specific needs of agentic systems, where latency, throughput, and efficient scheduling matter across many short or tool-heavy requests. -
15
RTP-LLM
Alibaba's high-performance LLM inference engine for diverse apps
RTP-LLM is an open-source large language model inference acceleration engine developed by Alibaba to provide high-performance serving infrastructure for modern LLM deployments. The system focuses on improving throughput, latency, and resource utilization when running large models in production environments. It achieves this by implementing optimized GPU kernels, batching strategies, and memory management techniques tailored for transformer inference workloads. ... -
16
Chitu
High-performance inference framework for large language models
Chitu is a high-performance inference engine designed to deploy and run large language models efficiently in production environments. The framework focuses on improving efficiency, flexibility, and scalability for organizations that need to run LLM inference workloads across different hardware platforms. It supports heterogeneous computing environments, including CPUs, GPUs, and various specialized AI accelerators, allowing models to run across a wide range of infrastructure configurations. ... -
17
gemma.cpp
lightweight, standalone C++ inference engine for Google's Gemma models
Gemma.cpp is a C++ implementation for running inference with Gemma models efficiently on CPUs and GPUs. Developed by Google, it allows running large language models (LLMs) like Gemma with minimal hardware, focusing on optimized performance and low latency. Gemma.cpp is intended for developers seeking to deploy LLMs in production environments without needing massive computational resources. -
18
WASTE
Run the full 2.78-trillion-parameter Kimi K3 model
WASTE is an embeddable C inference engine for running extremely large mixture-of-experts models when the weights exceed available RAM. It keeps the shared model trunk in memory and streams only the experts selected for each token from fast NVMe storage. A bounded cache reuses recently needed experts, while lookahead routing begins disk reads before the next layer requires them. -
19
HunyuanWorld-Voyager
RGBD video generation model conditioned on camera input
...The system jointly produces aligned RGB and depth video sequences, making it directly applicable to 3D reconstruction tasks. At its core, Voyager integrates a world-consistent video diffusion model with an efficient long-range world exploration engine powered by auto-regressive inference. To support training, the team built a scalable data engine that automatically curates large video datasets with camera pose estimation and metric depth prediction. As a result, Voyager delivers state-of-the-art performance on world exploration benchmarks while maintaining photometric, style, and 3D consistency. -
20
Mooncake
Mooncake is the serving platform for Kimi
...The platform was originally developed as part of the serving infrastructure for the Kimi large language model system. Its architecture centers on a high-performance transfer engine that provides unified data transfer across different storage and networking technologies. This engine enables efficient movement of tensors and model data across heterogeneous environments such as GPU memory, system memory, and distributed storage systems. Mooncake also introduces distributed key-value cache storage that allows inference systems to reuse previously computed attention states, significantly improving throughput in large-scale deployments. ... -
21
Cactus
Low-latency AI inference engine optimized for mobile devices
Cactus is a low-latency, energy-efficient AI inference framework designed specifically for mobile devices and wearables, enabling advanced machine learning capabilities directly on-device. It provides a full-stack architecture composed of an inference engine, a computation graph system, and highly optimized hardware kernels tailored for ARM-based processors. Cactus emphasizes efficient memory usage through techniques such as zero-copy computation graphs and quantized model formats, allowing large models to run within the constraints of mobile hardware. ... -
22
mllm
Fast Multimodal LLM on Mobile Devices
mllm is an open-source inference engine designed to run multimodal large language models efficiently on mobile devices and edge computing environments. The framework focuses on delivering high-performance AI inference in resource-constrained systems such as smartphones, embedded hardware, and lightweight computing platforms. Implemented primarily in C and C++, it is designed to operate with minimal external dependencies while taking advantage of hardware-specific acceleration technologies such as ARM NEON and x86 AVX2 instructions. ... -
23
SAM 3
Code for running inference and finetuning with SAM 3 model
SAM 3 (Segment Anything Model 3) is a unified foundation model for promptable segmentation in both images and videos, capable of detecting, segmenting, and tracking objects. It accepts both text prompts (open-vocabulary concepts like “red car” or “goalkeeper in white”) and visual prompts (points, boxes, masks) and returns high-quality masks, boxes, and scores for the requested concepts. Compared with SAM 2, SAM 3 introduces the ability to exhaustively segment all instances of an... -
24
Colibrì
Run GLM-5.2 (744B MoE) on a 25GB-RAM consumer machine
Colibri is a compact inference engine designed to run the 744-billion-parameter GLM-5.2 mixture-of-experts model on consumer hardware. It keeps the dense portion of the quantized model in memory while streaming routed experts from a large disk-based store as they are needed. The runtime is implemented in pure C, requires no Python or BLAS during inference, and can operate without a GPU. -
25
mistral.rs
Fast, flexible LLM inference
mistral.rs is a fast and flexible LLM inference engine implemented in Rust, designed to run and serve modern language models with an emphasis on performance and practical deployment. It provides multiple entry points for developers, including a CLI for running models locally and an HTTP server that exposes an OpenAI-compatible API surface for easy integration with existing clients.
