Python · MIT
ARahim3/mlx-dspark
Up to 4× faster LLM decoding on Apple Silicon, lossless. Native MLX port of DeepSeek's DSpark & z-lab's DFlash speculative decoding — Gemma-4, Qwen3.8, Muse-Glimmer, Nemotron, LFM2.5, Ornith-1.0, ternary Bonsai-27B.
★ 734
Stars
63 forks · 20 open issues
+6.0/wk
Growth · measured
Stars counted between two of our own snapshots roughly a week apart. The repository page gives the exact span.
25.9
Momentum / 100
see the breakdown below
4mo ago
Created
last push 33d ago
Star history
How this score was built
| Component | Weight | Earned | Why |
|---|---|---|---|
| Velocity | 55 | 5.1 | +6.0/wk, log-scaled against a 750/day ceiling |
| Acceleration | 15 | 0.0 | last week ran 0.77× the monthly pace |
| Freshness | 15 | 12.4 | last push 33d ago; decays to zero at 180 days |
| Scale | 15 | 8.3 | 734 stars, log-scaled against 150k — stops a 3-star repo topping the board |
| Momentum | 100 | 25.9 | sum of the above |
Detail
| Topics | apple-siliconclaude-codecodexdflash2dsparkinference-enginellm-inferencelocal-llmmacosmetalmlxpiqwen3-8speculative-decoding |
| Matched on | LLM & Inference (llm-inference) |
| Short window | +6 stars over 7 days = 0.9/day |
| Long window | +29 stars over 26 days = 1.1/day |
| Lifetime average | 5.71 stars/day over 129 days |
| First seen here | 2026-10-11 |
| Last read | 2d ago |