„Meituan“ atvėrė prieigą prie trilijono parametrų DI modelio, apmokyto naudojant kiniškus lustus

15 sources
  • „Meituan“ antradienį pristatė „LongCat-2.0“ – 1,6 trilijono parametrų atvirojo kodo modelį, apmokytą naudojant 50 000 kiniškų lustų.
  • Skirtingai nei „DeepSeek“ V4-pro, kurioje vietiniai lustai buvo naudojami tik išvadoms daryti, „LongCat-2.0“ naudojo kiniškus ASIC lustus tiek išankstiniam mokymui, tiek išvadoms, praneša „South China Morning Post“.
  • „Meituan“ teigia, kad modelis pagrindiniuose vertinimo testuose prilygsta arba pranoksta „Google“ „Gemini 3.1 Pro“, „OpenAI“ „GPT-5.5“ ir „Anthropic“ „Claude Opus“.
Sources (15)
  1. 1 China's Meituan says new AI model trained on domestic chips www.reuters.com
  2. 2 Meituan debuts China’s biggest AI model trained on local chips www.scmp.com
  3. 3 China's Meituan says new AI model trained on domestic chips - CNA www.channelnewsasia.com
  4. 4 China's Meituan says new AI model trained on domestic chips - AOL www.aol.com
  5. 5 China's Meituan Says New AI Model Trained on Domestic Chips money.usnews.com
  6. 6 Meituan Open Sources LongCat-2.0 Under MIT License www.opensourceforu.com
  7. 7 China's Meituan says new AI model trained on domestic chips - WTAQ wtaq.com
  8. 8 China's Meituan says new AI model trained on domestic chips techxplore.com
  9. 9 Meituan Just Dropped a Trillion-Parameter Model Trained on 60,000 ... medium.datadriveninvestor.com
  10. 10 Meituan open sources LongCat-2.0, the 1.6T, near-frontier agentic ... venturebeat.com
  11. 11 China's Meituan says new AI model trained on domestic chips economictimes.com
  12. 12 China's Meituan says new AI model trained on domestic chips www.thestar.com.my
  13. 13 Meituan: LongCat-2.0 - ZenMux zenmux.ai
  14. 14 Coin Bureau x.com
  15. 15 Meituan's LongCat-Next - Threads www.threads.com

Leave a Reply

Your email address will not be published. Required fields are marked *

Stay informed and not overwhelmed, subscribe now!