Frontier Models Show Wide Performance Gap in Latency Benchmarks

Artificial Analysis' latest latency measurements reveal stark differences in response speed across the current generation of frontier language models, with reasoning-optimized variants significantly lagging behind faster alternatives. The benchmarking platform conducted latency tests on four leading models—GPT-6.1 Sol, GPT-6 Astra, Claude Fable 5.1, and Claude Opus 5.5—under zero-shot conditions, exposing tradeoffs between reasoning capability and response speed that are reshaping deployment decisions across the industry.