{"version":1,"type":"story","url":"https://digestai.news/story/nace-ai-releases-drex-max-v1-a-27b-decision-model-that-runs-locally-on","json":"https://digestai.news/story/nace-ai-releases-drex-max-v1-a-27b-decision-model-that-runs-locally-on.json","markdown":"https://digestai.news/story/nace-ai-releases-drex-max-v1-a-27b-decision-model-that-runs-locally-on.md","slug":"nace-ai-releases-drex-max-v1-a-27b-decision-model-that-runs-locally-on","headline":"Nace AI releases Drex Max v1, a 27B decision model that runs locally on Apple Silicon","summary":"Nace AI has released Drex Max v1 as open weights, a 27B decision model that can run on local hardware. The model ships bf16 weights, a Q8_0 GGUF, and an affine Q4 checkpoint, and can be used via llama.cpp or Ollama. It supports routing, classification and policy checks on a single machine. The model has a 262,144-token window (16,384 by default) and can output up to 255 options per question. It runs on NVIDIA GPUs (H100 bf16 or L40S 48 GB 8‑bit) and on Apple Silicon Macs with 24 GB+ (affine Q4) or 48 GB (Q8_0 GGUF). Performance figures are self‑reported: 67.52 on the public Decision Index 0.4, 6.27 points ahead of Drex v2.0, 97.3 skill on InfographicVQA and 81.9 in the Tools area. The release is open‑weights, allowing developers to keep routing, classification and policy checks on a local machine, which could improve privacy and reduce latency.","keyPoints":["Drex Max v1 is a 27B decision model with 262,144-token window and up to 255 options per question","It runs on NVIDIA GPUs and Apple Silicon Macs, with open‑weights via bf16, Q8_0 GGUF and affine Q4","Self‑reported performance: 67.52 on Decision Index 0.4, 97.3 on InfographicVQA, 81.9 in Tools area"],"whyItMatters":"Open‑weight decision models that run locally on consumer hardware can reduce reliance on cloud services, improve privacy and lower latency for routing and policy tasks.","category":{"slug":"models","name":"Generative AI & Models","url":"https://digestai.news/category/models"},"entities":{"companies":["Nace.AI"],"models":["Drex Max v1"],"people":["Asif Razzaq"]},"firstPublishedAt":"2026-11-01T08:10:07Z","updatedAt":"2026-11-01T08:10:07Z","sourceCount":1,"hasPrimarySource":false,"sources":[{"outlet":"MarkTechPost","title":"Nace AI Open-Sources Drex Max v1: A 27B Decision Model That Runs Locally on Apple Silicon","url":"https://marktechpost.com/2026/11/01/nace-ai-open-sources-drex-max-v1-local-inference","publishedAt":"2026-11-01T08:10:07Z","type":"press","primary":false,"lead":true}],"sourceNotes":null,"discussions":[],"thread":null,"cite":{"text":"Digest AI, \"Nace AI releases Drex Max v1, a 27B decision model that runs locally on Apple Silicon\", 1 November 2026, https://digestai.news/story/nace-ai-releases-drex-max-v1-a-27b-decision-model-that-runs-locally-on","publisher":"Digest AI","title":"Nace AI releases Drex Max v1, a 27B decision model that runs locally on Apple Silicon","datePublished":"2026-11-01T08:10:07Z","url":"https://digestai.news/story/nace-ai-releases-drex-max-v1-a-27b-decision-model-that-runs-locally-on"},"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"}