What it does
Core capabilities at a glance
- Apple Silicon
- Gemma
- Gemma4
- Gemma4 26B A4B
- Gpgpu
- LLM Inference
- Local AI
- Macos
Deep dive
The full breakdown - performance, comparisons, and setup
turbo-fieldfare
turbo-fieldfare is a local inference server - Gemma 4 26B-A4B inference in ~2 GB of RAM on any M-series MacBook.
Overview
Gemma 4 26B-A4B inference in about 2 GB of RAM A custom Swift + Metal runtime for any Apple Silicon Mac, even the 8 GB ones.
Memory got expensive. So I gave a 26-billion-parameter model a ~2 GB budget.
TurboFieldfare runs the instruction-tuned Gemma 4 26B-A4B without loading the entire 14.3 GB model into memory. It keeps the shared 1.35 GB core and FP16 KV cache in memory, then streams only the experts needed for each token from SSD. This is what lets the model run on Macs with 8 GB of RAM.
The runtime, streaming installer, CLI, and native Mac app are written in Swift and Metal. TurboFieldfare is model-specific rather than a wrapper around MLX or llama.cpp. The curated experiment record summarizes 103 measured results across kernels, caching, I/O, prefill, and decode.
On the first run, Swift Package Manager downloads and builds the Swift packages required by the tokenizer. The complete release build includes the foreground Mac app and its sibling decode-service executable.
When the app opens, choose Download and let TurboFieldfare fetch and repack the pinned model (about 15 GB). Once it is ready, choose Load Model, type your prompt, and press Generate.
turbo-fieldfare is open-source, written primarily in Swift, with 5,523 GitHub stars under the Apache 2.0 license. The latest release is 0.4.1 (2026-08-07).
Key capabilities
From the project's documentation:
- An Apple Silicon Mac; the validated target is an 8 GB M2 MacBook Air
- macOS 26 with Metal 4
- Xcode 26 and Swift 6.2 or newer
- Enough free storage for the ~14.3 GB model installation
- An internet connection for the first model install
- Remote streaming repack into the .gturbo model format
How it fits a local-AI stack
turbo-fieldfare runs on your own hardware, so pair it with a model and a GPU sized to your needs. Use the VRAM calculator to pick a model that fits your card, and see what you can run for hardware guidance. Related local inference servers in the directory:
Sources
- Source code & docs: drumih/turbo-fieldfare
Stats from GitHub, 2026-08-10.
Frequently asked
Quick answers to common questions
What is turbo-fieldfare?
turbo-fieldfare is a inference-server tool for local AI workloads. Gemma 4 26B-A4B inference in ~2 GB of RAM on any M-series MacBook
Is turbo-fieldfare free and open source?
Yes, turbo-fieldfare has 5,523 GitHub stars and is licensed under Apache 2.0. You can self-host it for free on macos.
What platforms does turbo-fieldfare support?
turbo-fieldfare runs on macos.
What hardware do I need for turbo-fieldfare?
The hardware requirements depend on which models you run. Check our hardware directory for compatible GPUs and systems. turbo-fieldfare has 5,523 GitHub stars and an active community.
Does turbo-fieldfare support GPU acceleration?
turbo-fieldfare supports GPU acceleration via CUDA, Metal, or Vulkan depending on your platform. For the best performance, pair it with an NVIDIA RTX 4090 or 5090.
What are the best alternatives to turbo-fieldfare?
Popular alternatives include other inference-server tools in our directory. Browse our full collection at /tool for comparisons, community reviews, and benchmark data to find the right fit for your workflow.
How much does turbo-fieldfare cost?
turbo-fieldfare is free-open-source. It is completely free and open source to self-host.
Pairs well with
Complementary tools, models, and hardware
Comments coming soon
Configure NEXT_PUBLIC_GISCUS_REPO_ID and NEXT_PUBLIC_GISCUS_CATEGORY_ID at giscus.app to enable.