Traders see a tightly contested field for third place in Text Arena Math by end of September, with OpenAI and Anthropic at roughly 50% implied probability and a broad group of labs clustered nearby. Recent model releases have narrowed gaps in mathematical reasoning benchmarks, as OpenAI’s GPT-5 variants, Anthropic’s Claude 4.x series, and Google’s Gemini 3.x continue to trade incremental leads on math-specific categories while Chinese labs such as DeepSeek V4 post competitive MATH and HumanEval scores at far lower training costs. The proliferation of rapid updates, open-weight releases, and specialized post-training techniques has left no clear hierarchy beyond the top two, creating uncertainty about which additional labs will sustain the performance needed to rank third on the live arena leaderboard. Upcoming developer conferences and expected frontier drops through summer will likely serve as the next catalysts that shift these closely matched odds.