TurboFieldfare icon
TurboFieldfare icon

TurboFieldfare

Custom Mac runtime allows instruction-tuned 26-billion-parameter model inference in about 2 GB RAM by streaming only required model experts from SSD, with 1.35 GB core and cache in memory. Supports 8 GB Apple Silicon Macs with CLI and native app, measured in 103 scenarios.

TurboFieldfare screenshot 1

Cost / License

Application type

Platforms

  • Mac
Alternatives  
4likes
16alternatives
0articles

Features

Properties

  1.  Privacy focused
  2.  Local-First
  3.  AI-Powered

Features

  1.  No registration required
  2.  Works Offline
  3.  Ad-free
  4.  Dark Mode
  5.  Command line interface
  6.  No Tracking
  7.  Apple Silicon support
  8.  Local AI

 Tags

TurboFieldfare News & Activities

Highlights All activities

Recent activities

TurboFieldfare information

AlternativeTo Categories

AI Tools & Services, OS & Utilities, System & Hardware

GitHub repository

  •  6,687 Stars
  •  425 Forks
  •  49 Open Issues
  •   Updated  
View on GitHub

Popular alternatives

View all

Our users have written 1 comments and reviews about TurboFieldfare, and it has gotten 4 likes

TurboFieldfare was added to AlternativeTo by Paul on and this page was last updated .

Comments and Reviews

   
Top Positive Comment
Viktor Korobov
0

This is a really clever idea. It's impressive to see such a large model running on an 8 GB Mac without needing high-end hardware. I probably wouldn't use it every day, but I always enjoy projects that find creative ways to make AI more accessible.

Official Links

AppStores & Other Links

What is TurboFieldfare?

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.