Mference
Swift + Metal MoE inference for Apple Silicon: Qwen 3.6 35B at 23.5–29.3 tok/s decode with 2.20× faster long-prompt prefill on a 24 GB M5; Gemma 4 26B in ~2 GB, DeepSeek-V4-Flash 284B, Inkling-Small 276B, native Mac app, CLI, and OpenAI-compatible server.
NeelM0906/Mference
View on GitHub ↗Swift + Metal MoE inference for Apple Silicon: Qwen 3.6 35B at 23.5–29.3 tok/s decode with 2.20× faster long-prompt prefill on a 24 GB M5; Gemma 4 26B in ~2 GB, DeepSeek-V4-Flash 284B, Inkling-Small 276B, native Mac app, CLI, and OpenAI-compatible server.
Mference grew +3.4% in the last 30 days, ranks 134th of 159 in Web / Edge / On-Device Runtime, last commit 29 days ago.
Star trajectory
62 days tracked+3.4% in 30d, above the Web / Edge / On-Device Runtime median of +1.8%.
Last commit 29 days ago, staler than the median of 9d.
How it compares in Web / Edge / On-Device Runtime
Ranked 134th of 159 by stars.
- Language
- Swift
- License
- Other
- Contributors
- 5
- Forks
- 15
- Watchers
- 5
- Open issues
- 1
- Releases
- 1
- Created
- 2026-07-31
- Stars this week
- +0
You've seen where Mference stands today.
Get the moves before everyone else: the 5 Web / Edge / On-Device Runtime repos climbing fastest, every Tuesday.
Free weekly digest. Double opt-in, unsubscribe anytime.