Radar / AI infrastructure / llama.cpp v1.0.0
llama.cpp v1.0.0: the local-AI engine's first stable release

In today's Spotlight · Picked by 4 of 4 modelsA widely used tool for running AI on your own computer now promises more predictable updates.Like a trusted appliance maker promising new models still fit old plugs.
llama.cpp and ggml hit v1.0.0, promising semantic versioning so breaking C API changes only arrive with major bumps.
Why it matters
After 3.5 years as the engine under most local AI apps, llama.cpp now offers a stability contract: FP8 KV caches halve memory, 35 deprecated functions are gone, and app developers can finally pin a stable API.
What you could build with it
A small team shipping a desktop app that runs open models locally could pin llama.cpp 1.0 as its inference core and advertise a stable, long-support API surface, letting it ship 32k-context features on consumer laptops with half the KV-cache memory.
Does it hold up?
The 1.0 tag is a versioning promise, not a new feature set, so existing setups keep working; the real win is downstream apps can now build on a stable C API, and the FP8 KV cache measurably halves cache memory on long contexts.
Built with llama.cpp v1.0.0
- llama.cpp 1.0 saves 896 MiB in Qwen's 32k KV payloadarticle · Works out the FP8 KV cache saving on Qwen2.5-7B: a full 32k-token cache drops from 1,792 MiB to 896 MiB.
- llama.cpp 1.0 locks in a stable C APIarticle · Covers the 1.0 release notes: 35 deprecated functions removed, FP8 caches, and new model support including Kimi-K3.1.
- Pin llama.cpp releases on your serverarticle · Practical guide to pinning the new v1.0.0 tag (build b12359) so a local inference server stays reproducible.
- ggml-org/ggmlgithub · The tensor library underneath llama.cpp, also released as v1.0.0 the same day with FP8 support and fused MoE routing on Apple Metal.
Learn more
First spotted on github: source.
More AI infrastructure
vLLM v0.34.0: extensible KV cache by default, Model Runner V2 only, and 10 new modelsThe open inference engine's latest release spans 1,061 commits from 412 contributors, adds 10 models including Mistral…infra · JEV 0.86colibri: run 744B-parameter MoE models on consumer hardwarePure-C inference engine with zero dependencies that streams MoE experts from NVMe disk, keeping hot experts on GPU…infra · JEV 0.78claude-mem: persistent memory plugin that lets coding agents pick up where the last session left offPlugin that hooks five Claude Code lifecycle events, compresses session activity into a local SQLite + Chroma store…infra · JEV 0.74FreeToken: datacenter-scale MoE serving on desktop GPUsAn edge-native mixture-of-experts serving engine from UC Berkeley researchers that runs 290B+ parameter MoE models on…infra · JEV 0.74GBrain: Garry Tan's self-wiring memory layer for AI agentsY Combinator CEO Garry Tan's open-source, markdown-first knowledge layer that gives OpenClaw and Hermes agents…infra · JEV 0.71AWS Emulator: a local copy of AWS, written entirely by AI agentsAWS's free, open-source local emulator covering 200+ services from 438 API models, written in Rust, running offline on…infra · JEV 0.67
Get the week's best AI launches, plus 3 ideas worth building
One email every Saturday. Ranked by traction, not hype. Free.