
dmlx
Decodes faster and starts answering sooner on Apple Silicon, using an MTP drafter.
What it does better
Time to first token on a short prompt falls from 0.91 s to 0.28 s against omlx-mtp, with the same Qwen3.8-27B 4-bit target and MTP drafter on an M4 Pro with 48 GB
Benchmarks
- On Apple M4 Pro 48GB with Qwen3.8-27B (4-bit), it decodes at 43.2 tok/s, against 36.1 tok/s for omlx-mtp (+20%). Source Note: Non-thinking decode headline, greedy, with the same MTP drafter for both engines. The author says the raw artifacts record engine versions as unknown, so it is a recorded local result, not a portable guarantee.
Listing
Listed: Passed all seven checks.
Security review
Not reviewed yet.
Install
To install it, follow the project's own instructions on GitHub. The engine runs with your user's permissions. macOS doesn't sandbox it, so install only engines you trust.
Something wrong or out of date? Write to help@kalsa.io, or ask us to remove this page.