DeepSeek v4.1 flash runs at 23 seconds per token on a 2020 M1 Mac Mini
ajay-higgs · hn · 2026-09-12
A demo shows DeepSeek v4.1 flash running locally on a 2020 16GB M1 Mac Mini at a glacial 23 seconds per token — a striking illustration of how poorly older hardware handles the newest models.
More from Fun
- Letting "GPT 6 Astra" loose on an HF account to generate data goes wrong — realmrfakename · 2026-09-14
- VRChat real-time fluid rendering hits a wall: sparse grid isosurface too slow, only runs on a 5090 — Michael_Moroz_ · 2026-09-14
- Claude hilariously misreads a 'recreate' prompt, user shares the fail — felpix_ · 2026-09-14
- Dev chains a motion model with an image diffusion model to make glitchy datamosh animation — pixlpa · 2026-09-14
- He screenshotted a tweet and had Muse build a kids' bank with 3% weekly interest — armand_ruiz · 2026-09-14
- Anthropic Alignment Lead Amanda Askell Sparks Firestorm Over Remarks on Discrimination — SumitGup · 2026-09-14