Fourth Paradigm has developed two new foundation models, including Phai-A1, utilizing Huawei Ascend silicon and bypassing Nvidia hardware. Phai-A1 is an enterprise-focused model designed for coding and agent tasks, boasting a 1-million-token context window. Notably, it can perform inference on a single consumer-grade RTX 3090, demonstrating the capability of localized Chinese hardware. AI
IMPACT Demonstrates the viability of non-Nvidia hardware for training advanced AI models, potentially diversifying the AI infrastructure landscape.
RANK_REASON Frontier-lab model release with system card. [lever_c_demoted from frontier_release: ic=1 ai=1.0]
Read on Mastodon — mastodon.social →
AI-generated summary · Google Gemini · from 1 sources. How we write summaries →