Moonshot AI’s July 2026 launch of the 2.8-trillion-parameter Kimi K3 has anchored trader views on the market, with the model posting 56% on Humanity’s Last Exam using tools and 43.5% without—placing it among the stronger non-Anthropic entries on independent leaderboards. Frontier models overall have advanced rapidly on the 2,500-question expert benchmark, climbing from single-digit scores in early 2025 to top marks near 65% by late 2026, driven by improved reasoning chains, tool integration, and larger context windows. Kimi’s open-weight releases and agentic variants support continued iteration through year-end, though closing the gap to Claude Fable 5.1 or Opus 5.5 leaders will require further gains in long-horizon problem solving. Key catalysts ahead include any December model updates or verified tool-use evaluations that could shift the highest Kimi score across 50% or 55% thresholds.