**["The Light of Korean Models" Solar Pro4 Intelligent Index Hits 42 Points, But Still Can't Beat DeepSeek V4 Flash]** According to monitoring by Beating, South Korean AI company Upstage has released its flagship inference model, Solar Pro 4. In the latest evaluation by Artificial Analysis, its intelligent index surged from 14 points in the previous generation Solar Pro 3 to 42 points, with major improvements in Agent capabilities and long-context reasoning. Specifically, Terminal-Bench v2.1 rose from 12% to 57%, and long-context reasoning improved from 31 to 71. GDPval-AA v2 also increased from 498 Elo to 1277, surpassing the human baseline of 1000. Solar Pro 4 has become a representative model for South Korea's push into the global AI forefront. However, when compared to DeepSeek V4 Flash 0731, it still falls short. In the same Artificial Analysis intelligent index, DeepSeek scored 52 points, 10 points higher than Solar Pro 4. DeepSeek outputs approximately 122 tokens per second, while Solar Pro 4 outputs around 65 tokens, making the former nearly twice as fast. The price gap is even more significant. For every million tokens, Solar Pro 4 costs $0.30 for input and $1.20 for output, while DeepSeek costs only $0.14 and $0.28 respectively. Solar's output price is over four times higher than DeepSeek's. At least based on this comprehensive evaluation and API cost-performance ratio, "The Light of Korean Models" has yet to catch up to DeepSeek. [Original Link]