{"version":1,"type":"story","url":"https://digestai.news/story/omlx-0-7-4-adds-prefill-accelerator-using-a-second-mac-for-large-model","json":"https://digestai.news/story/omlx-0-7-4-adds-prefill-accelerator-using-a-second-mac-for-large-model.json","markdown":"https://digestai.news/story/omlx-0-7-4-adds-prefill-accelerator-using-a-second-mac-for-large-model.md","slug":"omlx-0-7-4-adds-prefill-accelerator-using-a-second-mac-for-large-model","headline":"oMLX 0.7.4 adds prefill accelerator using a second Mac for large models","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.","keyPoints":["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%"],"whyItMatters":"Enables running frontier-scale models on consumer Mac clusters without specialized interconnects, lowering the hardware barrier for local inference.","category":{"slug":"models","name":"Generative AI & Models","url":"https://digestai.news/category/models"},"entities":{"companies":["oMLX"],"models":["DeepSeek V4.1","Qwen3.8 Flash-Next","GLM-5.3","Ming-3.0-tts","Ming-3.0 Tiny","LFM3"],"people":["junmo-kim","qoli","yohann-bearzi","alan13367","scaryrawr","ianderse"]},"firstPublishedAt":"2026-11-04T13:00:03Z","updatedAt":"2026-11-04T13:00:03Z","sourceCount":1,"hasPrimarySource":true,"sources":[{"outlet":"github.com","title":"oMLX 0.7.4 can pair a fast mac as prefill accelerator","url":"https://github.com/jundot/omlx/releases/tag/v0.7.4.dev1","publishedAt":"2026-11-04T13:00:03Z","type":"primary","primary":true,"lead":true}],"sourceNotes":null,"discussions":[{"site":"Reddit","url":"https://www.reddit.com/r/LocalLLaMA/comments/1xnnl57/omlx_074_can_pair_a_fast_mac_as_prefill/","points":null}],"thread":null,"cite":{"text":"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","publisher":"Digest AI","title":"oMLX 0.7.4 adds prefill accelerator using a second Mac for large models","datePublished":"2026-11-04T13:00:03Z","url":"https://digestai.news/story/omlx-0-7-4-adds-prefill-accelerator-using-a-second-mac-for-large-model"},"generatedBy":"Written by Digest AI's editorial model from the linked sources; the sources are the record.","license":"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"}