jinfer brings native AI inference to the JVM, matching llama.cpp on CPU with zero Python

mukel90 · reddit · 2026-09-15

Developer mukel90 (author of llama3.java and gemma4.java) released jinfer, an Apache 2.0 open-source inference engine built for the JVM from first principles: chat, vision, audio transcription, embeddings, reranking, and TTS—with no Python runtime, ONNX, or sidecar process. The stack includes toknroll (pure-Java zero-dependency tokenizers), native gguf/safetensors read/write, jam (quantized matmul routines with Vector API + optional native backend, competitive with llama.cpp on CPU), and jota (a tensor API targeting Java, C, CUDA, HIP, Metal, OpenCL, and Mojo). It integrates with Spring AI and LangChain4j and supports GraalVM Native Image. It's an early release: CPU is the main target today, GPU support is in progress. Benchmarks and runnable examples at qxotic.ai.

Original post →

More from Infra

Infra channel →