Hy4 Flash reads 1M tokens in 44 min on Mac Studio M3 Ultra, 81 min on DGX Spark
AIFlow_ML · x · 2026-11-21
Per @Razumasu's test, Hy4 Flash read a 1M-token prompt in 44 minutes on a Mac Studio M3 Ultra and 81 minutes on a DGX Spark. After that one slow read, cache-based follow-ups took only 2.8s on the Mac and 3.4s on the Spark.
The takeaway: a "read once overnight, then query instantly all day" workflow is practical for large documents.
Related event: Local benchmarks reveal Hunyuan Hy4 Flash speed and long-prompt load times(2 posts)→
More from Infra
- local-ai-registry: one battle-tested local LLM recipe per GPU, with evidence attached — LiTianleli · 2026-11-21
- Sous Labs open-sources Songe, a local fine-tuning engine that trains on your own conversations — alexcovo_eth · 2026-11-21
- AI supply chain: everyone 50-100% sold out for 2027, capex set to keep rising through 2028 — sudoraohacker · 2026-11-21
- Agent auto-optimizes new model 7x to 210 tok/s on Apple Silicon with zero guidance — simonguozirui · 2026-11-21
- GCMP update: LoRA-quality at 2.9M params, serves 256 adapters on one A40 with zero throughput drop — textclf · 2026-11-21
- Four DGX Sparks, one week of tuning: 26.85 to 56.71 tok/s — AIFlow_ML · 2026-11-21