VIDRAFT's Darwin-180B-RSI tops MMLU-Pro and GPQA Diamond

Korean startup VIDRAFT released Darwin-180B-RSI, a 180B-parameter open-weight model. Per aggregator AIhoster, it ranks first on MMLU-Pro (88.1%) and GPQA Diamond (94.4%), beating models with far more parameters.

• MMLU-Pro: 88.1% across all 12,032 test items
• GPQA Diamond: 94.4% with majority voting over up to 16 samples
• Thinking budget up to 131,072 tokens, bf16 precision
• Weights published at FINAL-Bench/Darwin-180B-RSI on Hugging Face