Claude Opus 5.5 leads ArtificialWatch’s combined benchmark rankings

Claude Opus 5.5 leads the ArtificialWatch Bench Index with 98.1, followed by Gemini 4 Carbon at 95 and GPT-6 Astra at 92.6. The ranking, refreshed on 2026-11-06, covers 45 ranked models across 6 public benchmark boards. Google’s runner-up has been announced but is not publicly available; its scores reflect early-access versions.

The index combines LMArena, LiveBench, SimpleBench, LLM Stats, Vals AI and Artificial Analysis. Each model’s best-performing variant on each board is converted into a percentile rank, and the median of those percentiles determines its composite score. Search, grounding, image and video product modes are excluded. Models need results from 2+ boards to qualify, while boards scoring fewer than 5 of the listed models do not contribute to the index.

The methodology combines community preferences, reasoning tests, live task suites and specialist evaluations without averaging incompatible raw scores. Percentiles are calculated relative to the models in ArtificialWatch’s table, not each benchmark’s full roster. The table is rebuilt daily, with source links and fetch dates attached to raw scores; no missing results are estimated.