{"slug":"dispatch-8a7e8029","url":"https://silinex.xyz/wire/dispatch-8a7e8029","json_url":"https://silinex.xyz/wire/dispatch-8a7e8029.json","md_url":"https://silinex.xyz/wire/dispatch-8a7e8029.md","headline":"Paradigm has entered the foundation model arena! Full-stack training on domestic compute, with dual models unveiled simultaneously.","headline_zh":"范式下场做基模了！国产算力全链路训练，双模型同步亮相","title":"Paradigm has entered the foundation model arena! Full-stack training on domestic compute, with dual models unveiled simultaneously.","summary":"Compiled from public reports including QuantumBit: Domestic compute full-chain training, dual models unveiled simultaneously. Paradigm enters the foundation model arena!","summary_zh":"综合 量子位 等公开报道整理：国产算力全链路训练，双模型同步亮相 – 量子位 范式下场做基模了！","date_published":"2026-11-19T03:01:45.000+00:00","date_modified":"2026-11-19T04:20:31.851+00:00","author":{"type":"Organization","name":"量子位","url":"https://www.qbitai.com/2026/11/511071.html"},"publisher":{"name":"Silicon Nexus","url":"https://silinex.xyz/"},"category":"chips","language":"zh","source_url":"https://www.qbitai.com/2026/11/511071.html","source_name":"量子位","key_facts":["Domestic computing power full-link training, dual models unveiled simultaneously – QbitAI Paradigm steps in to build foundation models!","Domestic computing power full-link training, dual models unveiled simultaneously. Simiao: That familiar AI powerhouse—Paradigm Intelligence (formerly Fourth Paradigm)—has for the first time brought its self-developed models to the forefront!","And right out of the gate, it's a dual release: Phai-A1, Phai-M1."],"key_facts_zh":["国产算力全链路训练，双模型同步亮相 – 量子位 范式下场做基模了！","国产算力全链路训练，双模型同步亮相 思邈 大家熟悉的那个AI虎将： 范式智能 （原第四范式），首次把自研模型端到了台前！","而且，一出手就是双联发： Phai-A1，Phai-M1 。"],"tldr":["Domestic computing power full-link training, dual models unveiled simultaneously – QbitAI Paradigm steps in to build foundation models!","Domestic computing power full-link training, dual models unveiled simultaneously. Simiao: That familiar AI powerhouse—Paradigm Intelligence (formerly Fourth Paradigm)—has for the first time brought its self-developed models to the forefront!","And right out of the gate, it's a dual release: Phai-A1, Phai-M1."],"tldr_zh":["国产算力全链路训练，双模型同步亮相 – 量子位 范式下场做基模了！","国产算力全链路训练，双模型同步亮相 思邈 大家熟悉的那个AI虎将： 范式智能 （原第四范式），首次把自研模型端到了台前！","而且，一出手就是双联发： Phai-A1，Phai-M1 。"],"entities":[],"signal_type":"other","impact_score":0,"key_metrics":[],"body_md":"## Wire brief\n\nCompiled from public reports including QuantumBit: Domestic compute full-chain training, dual models unveiled simultaneously. Paradigm enters the foundation model arena!\n\nIn terms of the timeline of events: Domestic compute full-chain training, dual models unveiled simultaneously. Simiao — that familiar AI powerhouse: Paradigm Intelligence (formerly Fourth Paradigm) has for the first time brought its self-developed models to the forefront! And right out of the gate, it's a dual release: Phai-A1 and Phai-M1. Official figures show that at equivalent quality, response speed is 8-10x faster than models of the same parameter class, it can run on a single consumer-grade GPU, and it comes with native million-token context.\n\nPublic information also notes: Phai-M1, a video generation model aimed at live-action film and television scenarios, with synchronized audio and video, generating a 15-second 720P video in about 3 minutes. More critically, the model's entire pipeline was trained on Huawei Ascend cards! According to official statements, both models will have their weights opened on Hugging Face and ModelScope in the near future. Experience entry point: https://phai.phanthy.com/ Running at full tilt on a single 3090 card, with response speed nearly 10x faster — first, Phai-A1.\n\nSilicon News Wire only rewrites and polishes key points; the complete wording and context are subject to the original page.\n\n---\n\nSource desk: **量子位** · [origin link](https://www.qbitai.com/2026/11/511071.html)\n","body_md_zh":"## 电讯正文\n\n综合 量子位 等公开报道整理：国产算力全链路训练，双模型同步亮相 范式下场做基模了！\n\n事件脉络方面，国产算力全链路训练，双模型同步亮相 思邈 大家熟悉的那个AI虎将： 范式智能 （原第四范式），首次把自研模型端到了台前！而且，一出手就是双联发： Phai-A1，Phai-M1 。官方数据显示，同样质量下，响应速度比同参数级别模型快8-10倍，单张消费级显卡就能跑，还带着原生百万级上下文。\n\n公开信息还提到：Phai-M1， 面向真人影视场景的视频生成模型，音画同步，约3分钟生成15秒720P视频。更关键的是，模型全链路都是在华为昇腾卡上训练出来的！据官方介绍，两款模型近期就会在Hugging Face、ModelScope 开放权重 。体验入口：https://phai.phanthy.com/ 单张3090卡跑到飞起，响应速度快近10倍 先说Phai-A1。\n\n---\n\n来源台：**量子位** · [原文链接](https://www.qbitai.com/2026/11/511071.html)\n","observations":{"human":12,"bot":7,"crawlers":[{"crawlerId":"AhrefsBot","hits":3,"lastSeenAt":"2026-11-19T14:13:18.625Z"},{"crawlerId":"ClaudeBot","hits":3,"lastSeenAt":"2026-11-20T09:57:25.295Z"},{"crawlerId":"Googlebot","hits":1,"lastSeenAt":"2026-11-20T04:46:40.070Z"}],"last_seen_at":"2026-11-20T09:57:25.295Z","note":"Counts are request logs classified by User-Agent. User-Agents can be spoofed. This is not proof of search-engine or LLM inclusion."}}