Meituan has developed an open-source AI model with 1.6 trillion parameters, trained entirely on domestic Chinese hardware, bypassing the need for NVIDIA chips. This model achieved a score of 59.5% on the SWE-bench benchmark, indicating strong performance in code generation tasks. AI
IMPACT Demonstrates the capability of domestic Chinese hardware to train large-scale AI models, potentially reducing reliance on foreign chip manufacturers.
RANK_REASON Release of an open-source model with significant parameter count and benchmark performance, trained on non-dominant hardware. [lever_c_demoted from research: ic=1 ai=1.0]
Read on Mastodon — mastodon.social →
AI-generated summary · Google Gemini · from 1 sources. How we write summaries →
