GLM-5.3 beats Anthropic/OpenAI models at a fifth of the cost

GLM-5.3 (open-weight) beat Anthropic/OpenAI models – for 1/5 the cost

The Ed-o-meter leaderboard pits 17 leading LLMs against 28 real-world tasks—coding, data, realworld, security, and tool-use—using identical prompts and deterministic grading. GLM-5.3 is the first to clear all five corners at 100%, with a 9.3 rubric score and a lap cost of just $0.28, about a fifth of GPT-5.5's $1.43. GPT-5.5 is faster (13.2s vs 16.3s TTFT) but scores 89% on realworld. The GPT-5.6 line fails security (33–50% pass) due to jailbreak canaries, while Claude models refuse benign tasks. Notable: Kimi-K3 tops the rubric at 9.5, and GPT-5.6-Luna is the cheapest workhorse at $0.064 per lap.