GLM 5.1 beats every model except Opus in agent tasks—and costs 1/3 as much

You know how AI companies love to game benchmarks? GLM 5.1 looked suspicious at first—another model that might just be optimized for test scores. So I actually tested it on real agent tasks (like OpenClaw) to see if it was the real deal. Turns out it's genuinely impressive. GLM 5.1 performs at Claude Opus 4.6 level—basically the gold standard—but costs only ~$0.4 per run instead of ~$1.2. That's a third of the price for the same quality. It crushes every other model in the benchmark too. This is actually a big deal for anyone building AI agents, because suddenly you can get top-tier performance without burning through your budget. The cost-effectiveness frontier just shifted hard. Worth checking out if you're serious about agents.