Solar Pro 2 Breaks Into Global Frontier AI

Upstage has earned global recognition with its latest proprietary large language model (LLM), Solar Pro 2, which was recently named one of the top 10 “Frontier Models” by leading AI benchmark organization Artificial Analysis. It is the only model developed in Korea to be included in the index, placing Upstage alongside industry leaders such as OpenAI, Google, Meta, and Anthropic.

Artificial Analysis’ Intelligence Index is regarded as one of the most comprehensive public evaluations of language model performance. The index assesses over 200 models using seven key benchmarks—including reasoning, coding, general knowledge, and advanced math—to generate a composite score that reflects multiple dimensions of intelligence.

Solar Pro 2 achieved a score of 58, outperforming widely used models such as GPT-4.1 (53), GPT-4o (41), and LLaMA 4 Maverick (51). It also surpassed notable newcomers such as Kimi-K2 (57.59) and DeepSeek V3 (53)—a significant achievement for a model with just 31 billion parameters. In contrast, many of the models it surpassed or matched operate at 100B+ scale, with the top-ranked Grok-4 reportedly reaching 1.7 trillion parameters.

The recognition highlights Solar Pro 2 as a rare example of a compact model delivering near-frontier performance. In addition to outperforming larger models in reasoning-heavy tasks, it also ranked highly in cost-efficiency metrics, making it a viable option for enterprise use where performance and cost must be carefully balanced.

While GPT-4.1 remains a dominant force in real-world applications, especially across enterprise AI deployments, Solar Pro 2’s higher score suggests stronger capabilities in complex reasoning and knowledge synthesis—functions increasingly central to the next generation of AI systems.

The inclusion of Upstage in the “Frontier Model” category marks a milestone for the Korean AI ecosystem and reflects a growing diversification in global AI leadership. Upstage is now listed alongside OpenAI, Google, Meta, and Anthropic on Artificial Analysis’ official rankings.

Looking ahead, Upstage is currently developing a 100B+ parameter model aimed at pushing further into the upper tier of the Intelligence Index.

“Solar Pro 2 shows that globally competitive foundation models can be built outside of Silicon Valley,” said Sung Kim, CEO of Upstage. “We’re proud to bring Korean innovation to the global stage, and we’ll continue to develop scalable, high-performance models that help shape the future of AI worldwide.”