DeepSeek-V2 Outperforms Llama3 in Chinese Benchmarks at 1/10 the Cost, Sparking Heated Discussions
DeepSeek team released V2 open-source model, excelling in Chinese benchmarks especially in math and code generation, surpassing Meta's Llama3 while costing only 1/10 to train. The breakthrough ignited discussions on X platform's Chinese community with over 500,000 views.