# oMLX 0.7.4 adds prefill accelerator using a second Mac for large models

Digest AI · Generative AI & Models · published 2026-11-04T13:00:03Z

Canonical: https://digestai.news/story/omlx-0-7-4-adds-prefill-accelerator-using-a-second-mac-for-large-model

## Summary

The oMLX project released version 0.7.4.dev1, a development build that lets a second Mac with a faster GPU accelerate the prefill stage for large models that only fit on a high-memory Mac. The project reports prefill throughput improvements of 43-48% for DeepSeek V4.1 on an M5 Ultra and 16% faster batched decode for Qwen3.8-27B on an M5 Max. New TTS models Ming-3.0-tts and Ming-3.0 Tiny show real-time factor gains of 52% and 68% respectively. The release also adds integrations for Codex, Codex CLI and GitHub Copilot, German localization across the app, and several bug fixes for tool-call ordering, prompt caching and memory reporting. This is a development release; the project advises using 0.7.3 for stability.

## Key points

- Prefill accelerator pairs a fast-GPU Mac with a high-memory Mac over Thunderbolt or 5GbE
- DeepSeek V4.1 prefill 43-48% faster on M5 Ultra; Qwen3.8-27B batched decode +16% on M5 Max
- New TTS models Ming-3.0-tts and Tiny cut real-time factor by 52% and 68%

## Why it matters

Enables running frontier-scale models on consumer Mac clusters without specialized interconnects, lowering the hardware barrier for local inference.

## Sources

1. [oMLX 0.7.4 can pair a fast mac as prefill accelerator](https://github.com/jundot/omlx/releases/tag/v0.7.4.dev1) (github.com, 2026-11-04, primary source)

## Cite

Digest AI, "oMLX 0.7.4 adds prefill accelerator using a second Mac for large models", 4 November 2026, https://digestai.news/story/omlx-0-7-4-adds-prefill-accelerator-using-a-second-mac-for-large-model

---

Written by Digest AI's editorial model from the linked sources; the sources are the record. Headlines, digests and key points are written by Digest AI and may be quoted with a link to the story page. Linked articles belong to their publishers. Terms: https://digestai.news/terms#reuse
JSON: https://digestai.news/story/omlx-0-7-4-adds-prefill-accelerator-using-a-second-mac-for-large-model.json