SciUniverse puts models in a real wet lab, and Fable 5.1 leads GPT-6 Astra 45.3% to 32.5%.

C5R's benchmark, Level 1 published September 24, has 92 tasks in 17 families across chemistry, biology and materials, where models control instruments and instruct human operators at C5R's Facility-0. One task synthesizes N-benzyl-4-methylbenzamide and confirms it by LC-MS. Pass@1: Fable 5.1 xhigh 45.3% at $40.61 per attempt, GPT-6 Astra 32.5% at $52.37, Opus 5 30.5%, Grok 4.6 26.2%, Gemini 3.8 Flash 14.6%, GPT-5.6 Sol 9.4%. Fable wins on both score and cost, which the widely circulated demo videos didn't show because they only ran Astra.