Claude Opus 5 Tops Artificial Analysis Intelligence Leaderboard at 61

Anthropic’s Claude Opus 5, running in Adaptive Reasoning at Max Effort, now sits at the top of the Artificial Analysis Intelligence Index with a score of 61 across the 170 models the benchmark has evaluated. Anthropic dominates the upper ranks: four of the top five slots are Claude variants, including Opus 5 at Xhigh Effort (60), Fable 5 at Max Effort (60), and Opus 5 at High Effort (59). The only outside entry in that group is OpenAI’s GPT-5.6 Sol (max) at 59. Opus 5 also leads the field of 126 reasoning models, which spend extended compute working through problems before answering.

The open-weights picture is a different race. GLM-5.2 (max) is the strongest open model at 51, ahead of MiniMax-M3 and DeepSeek V4 Pro (both 44), and 94 of the 170 ranked models ship open weights — a reminder that the frontier remains proprietary but the gap is roughly ten index points. Intelligence is only one axis: on speed, Mercury 2 clears 901.6 tokens per second, Nova Micro and a couple of rivals bottom out pricing at $0.03 per million tokens, and Gemini 2.5 Flash-Lite posts the lowest time-to-first-token at 0.33 seconds.

The leaderboard’s value is in how it separates these dimensions — quality, cost, throughput, latency, and context window — rather than collapsing them into a single number, with performance metrics measured directly across 586 models using standardized prompts. Opus 5’s placement signals that adaptive, effort-scaled reasoning is currently the shortest path to top-tier benchmark intelligence, at least until competitors close in.