Does an open-weights model surpass GPT-4 on MMLU by EOY 2026?

Resolves Yes if any model with weights released under an open license (Apache 2.0, MIT, or comparable) scores higher than GPT-4 on the published MMLU benchmark by December 31, 2026. Resolves based on the standard 5-shot MMLU evaluation as reported on the Hugging Face Open LLM Leaderboard or an equivalent third-party eval.