The 80B Qwen3-Coder-Next on a 16 GB Mac, about 13 tok/s, fully offline. A mixture-of-experts inference engine for Apple silicon.
Rust
| Name | Last commit message | Last commit date |
|---|---|---|
| .. | ||
| avx.rs 11.8 KB | ||
| erf.rs 9.4 KB | ||
| features.rs 2.1 KB | ||
| kernels.rs 8.2 KB | ||
| mod.rs 10.9 KB | ||
| neon.rs 20.4 KB | ||
| simd128.rs 8.1 KB | ||