Recent Text Arena lab rankings as of late September 2026 place Google and Anthropic at the top overall, with Meta, Alibaba’s Qwen3.8 series, Moonshot’s Kimi models, Xiaomi’s Mimo, and Z.ai’s GLM close behind in aggregate scores. Math-specific performance shows particular strength from Chinese labs, including Alibaba Qwen variants and Moonshot Kimi entries on benchmarks like FrontierMath and HMMT, alongside competitive results from DeepSeek and Meta’s Muse Spark models. With the market resolving at month-end and many outcomes clustered near even implied probabilities, traders appear focused on narrow gaps in recent blind voting data, potential late-October model updates or fine-tunes, and differentiation in reasoning depth versus general capability. This creates a tightly contested field where incremental benchmark gains or new releases could shift third-place positioning.