Claude Opus 5, with adaptive performance and maximum effort, currently occupies the first position in the Artificial Analysis Intelligence Index, achieving a score of 61 out of 170 models analyzed, as reported by artificialanalysis.ai.
This model stands out in a range of metrics that includes intelligence, price, response speed (tokens per second), latency (time for the first token), total response time, and context window size. Among the top models, Claude Opus 5 (Adaptive Reasoning, Xhigh Effort) and Claude Fable 5 (Adaptive Reasoning, Max Effort, Opus 4.8 Fallback) share the second and third positions with 60 points each.
Regarding speed, Mercury 2 leads with 901.6 tokens per second, followed by HyperNova 60B 2605 (439.5 t/s) and Gemini 3.5 Flash-Lite (436.5 t/s). In terms of financial accessibility, Nova Micro is the cheapest, with a price of $0.03 per 1M of tokens, followed by Sarvam 30B (high) and Gemma 4 E4B (No-reasoning), both also at $0.03.
In terms of latency, Gemini 2.5 Flash-Lite (No-reasoning) has the shortest time for the first token, with only 0.33s, followed by Command A+ (0.42s) and North Mini Code (0.49s).
Among open-weight models, GLM-5.2 (max) is the best rated with a score of 51, out of 94 open models in 170 evaluated. The top open-weight models, according to the Intelligence Index score, are GLM-5.2 (max), MiniMax-M3, and DeepSeek V4 Pro (Reasoning, Maximum Effort), with scores of 51, 44, and 44 respectively.
Claude Opus 5 (Adaptive Reasoning, Maximum Effort) also leads among the 126 reasoning models, which use extended thinking to solve complex problems before providing answers, with a score of 61.
These comparisons cover various dimensions, including intelligence quality, prices, token production speed, latency, total response time, and context window size. The performance metrics are measured directly using standardized prompts in 586 models.
By clicking on any model name or line in the charts, you can view its dedicated page with detailed metrics and direct comparisons with similar models. You can also use the model selector to customize which models appear in each chart and view the ranking.