Claude tops modelgrep intelligence rankings

Modelgrep’s rankings, updated November 2026, place Claude Opus 5.5 at the top for intelligence with 59.1 intelligence. Mercury 2.5 leads output speed at 874 tokens/sec, while GLM 5.2 has the lowest listed input price at $0.0060 per million input tokens. The rankings separate overall capability from cost, responsiveness and performance on specific tasks.

Claude Opus 5.5 also leads the reasoning, vision, writing, retrieval-augmented generation and data visualization categories. GPT-6 Astra takes the leading positions for science, SVG generation, UI components, game development and full-stack app building. GPT-5.6 Sol ranks highest for agents, while Muse Spark 1.3 leads design and frontend work.

For local deployment and open-weight models, Kimi K3.1 Pro holds the top position. Kimi K3 leads long-context reasoning, while Grok 4.20 Multi-Agent has the largest context window. Command A+ leads time to first token, GLM 5.2 tops tool calling, and MiniMax M3 leads instruction following. Modelgrep describes its comparisons as drawing on independent benchmarks, live speed measurements and pricing data.