Grok is pretty well performing on discovery / search aggregation so for search-heavy tasks it's been great for me.
I did some experiments: let Grok make a good list of reference materials and then let either GPT or Claude work that inventory in xhigh. This went well but I lack the volume to have definitive answers on this.
GLM-5.2 has been equal/near-equal to Opus-4.6/4.7 in analytic skills for me (and often outperforming 4.8/5). Kimi K3 outperforms Opus on expensive long-running tasks.
Grok is pretty well performing on discovery / search aggregation so for search-heavy tasks it's been great for me.
I did some experiments: let Grok make a good list of reference materials and then let either GPT or Claude work that inventory in xhigh. This went well but I lack the volume to have definitive answers on this.
GLM-5.2 has been equal/near-equal to Opus-4.6/4.7 in analytic skills for me (and often outperforming 4.8/5). Kimi K3 outperforms Opus on expensive long-running tasks.