Solar Open 2 language model boasts 1M-token context window and strong agentic skills

Researchers have introduced Solar Open 2, a 250 billion parameter Mixture-of-Experts language model designed for long-horizon agentic tasks. This model features a 1 million token context window achieved through a hybrid attention mechanism and is trained efficiently using a stronger starting point and curated high-value data. Solar Open 2 demonstrates leading performance on several English benchmarks, including MMLU-Pro and the APEX-Agents suite, and shows competitive results against other top open-weight models. It also achieves top scores on Korean benchmarks and an in-house Korean officework-agent benchmark, outperforming larger models. AI

IMPACT Sets new SOTA on agentic benchmarks and demonstrates efficient training for large context windows.

RANK_REASON Technical report detailing a new language model release with benchmark results. [lever_c_demoted from research: ic=1 ai=1.0]

Read on arXiv cs.CL →

• APEX-Agents
• DeepSeek-V4-Flash
• DeepSeek-V4-Pro
• LiveCodeBench
• MiMo-V2.5
• Mixture-of-Experts
• MMLU-Pro
• Solar Open 1
• Solar Open 2

• model release
• paper

AI-generated summary · Google Gemini · from 1 sources. How we write summaries →