Running 104GB Qwen3.8-Flash-Next on 48GB Mac with at ~12 tok/s

developer-tools / Show HN

SCORE

Run Qwen3.8-Flash-Next, a 104GB model, on 48GB Mac hardware with inference speeds around 12 tokens per second. Implements slot-based streaming for efficient large language model execution on resource-constrained devices.

VISIT PRODUCT →

Sources (1)

Show HN

https://news.ycombinator.com/item?id=49524447

65 PTS

Score Breakdown

Traction

raw 115.00 · weight 35%

24.1pts

Novelty

9.2 days old · weight 20%

19.5pts

Source diversity

1 source · weight 10%

3.3pts

AI quality

raw 62.00 · weight 35%

21.7pts

Technically impressive optimization achieving 12 tok/s on constrained hardware, but niche use case with unclear practical value proposition and minimal adoption signals.

Final Score69/100