Solar Open 100B (Reasoning) is Upstage's large open-weight reasoning model, released in December 2025. Its intelligence index of 21.7 is supported by a GPQA of 0.657 and HLE of 0.092, which are reasonable for a model at this tier. The agentic index of 25.3 and TAU2 of 0.482 indicate moderate multi-step task capability, and IFBench of 0.577 reflects functional instruction following. LCR of 0.360 is adequate for lighter document tasks.

For businesses, this model is best suited to content generation, summarisation, and structured output tasks where its large parameter count may offer some depth advantage. Coding capability (10.5) is limited, and the terminal bench score is very low, ruling it out for software engineering or agentic coding workflows. As an open-weight model, self-hosting is a key advantage for teams with data privacy requirements.

No pricing data is available, so cost-performance assessment depends on infrastructure costs. Teams considering Solar Open 100B (Reasoning) for on-premise or private cloud deployment will find it a functional option for content workloads, but should benchmark it against smaller, more capable models before committing to the infrastructure overhead of a 100B parameter deployment.

Assessed June 6, 2026

Rankings consider pricing, capabilities, benchmarks, and real-world applicability and are refreshed as new models launch. Feedback?