Anthropic holds the market lead at 58% implied probability due to its recent Claude Fable 5 and Opus 5 releases, which currently top LiveBench's contamination-free coding category with scores around 86%. These models demonstrate strong agentic coding and reasoning capabilities on live tasks, outpacing OpenAI's GPT-5.6 variants that sit near 82% on the same benchmark. With probabilities clustered around 50% for several contenders and OpenAI at 29.5%, traders see a tight race driven by rapid iteration cycles from multiple labs. Key swing factors over the next ten weeks include any new model drops, LiveBench task refreshes, or demonstrated gains in coding-specific benchmarks before the October cutoff, underscoring how quickly frontier large language model performance can shift.