inference-engine

Hand-written NVFP4 W4A16 CUDA kernels for Volta

hummingbird is a lightweight, zero dependency runtime for massive open source Mixture of Experts (MoE) language models. It unifies SSD, RAM, and VRAM into a single intelligent memory hierarchy, enabling inference of models like GPT-OSS 120B, GLM, DeepSeek, Qwen, and more on consumer hardware

[⛔️ DEPRECATED] Friendli: the fastest serving engine for generative AI

From-scratch C++23/CUDA inference engine for the NVIDIA RTX 5090 (sm_120a). The best single-GPU backend for agentic AI: tool calling, long-context loops, reasoning and concurrent sub-agents. Decode beats llama.cpp b9976 by 42-48% on dense GGUF (measured 2026-07-12), at-or-ahead of vLLM on NVFP4. 100% written by Claude Code.

Comprehensive, scalable ML inference architecture using Amazon EKS, leveraging Graviton processors for cost-effective CPU-based inference and GPU instances for accelerated inference. Guidance provides a complete end-to-end platform for deploying LLMs with agentic AI capabilities, including RAG and MCP

AI-Inference-Managed-by-AI: Go binary for managing AI inference on edge devices

Run NVIDIA's tri-modal Nemotron Omni (text+vision+audio) entirely on Apple Silicon — pure MLX runtime, parity-verified against NVIDIA's reference