Open source doubles M5 Ultra MLX token prefill in just one week via Flash-Next
TheMoonMidas · x · 2026-09-29
Developer viticci benchmarked recent oMLX upstream builds and found MLX performance on the M5 Ultra has improved dramatically in a single week: token prefill has essentially doubled across the board thanks to Flash-Next. Maintainer jundotkim reshared the results, noting most of the gains came from open source contributors rather than himself, and thanked him for measuring properly. The tester says he'll need to update his review soon.
More from Infra
- Sarvam AI Launches Token Factory: India-Hosted Inference for Open Models — itsOmSarraf_ · 2026-11-16
- BIS Issues Interim Final Rule Addressing Remote Access to Advanced AI Chips — apples_jimmy · 2026-11-16
- Jailbroken PS4 runs a 9B open-source model at up to 43 tok/s by talking directly to its 2013-era GPU — AIFlow_ML · 2026-11-16
- Mobile AI Factory: agents auto-build on-device SLMs with 19x less GPU time — pmttyji · 2026-11-16
- Stratechery's Fictional Obituary: Imagining Jensen Huang's 2026 Exit From Nvidia — Gregory_C_Allen · 2026-11-16
- vLLM serves DeepSeek V4.1-Flash at 954K tokens/s per GPU, $38.96/GPU-hour on GB300 NVL72 — bookwormengr · 2026-11-16