GLM-5.3 (open-weight) beat Anthropic/OpenAI models – for 1/5 the cost
Summary
The Ed-o-meter benchmark compares 17 LLMs across 28 real-world tasks using a uniform setup, identifying glm-5.3 as a cost-effective all-five-corners winner at 100% and $0.28 per lap. It also discusses latency trade-offs, safety filters, and alternative models like gpt-5.5 and luna for faster or cheaper use cases, with Featherbench as the open-source harness behind the results.