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

The recent release of GLM-5.3 has made waves in the AI/ML community after outperforming renowned models from Anthropic and OpenAI at a fraction of the cost, demonstrating its potential as a leading model for practical applications. In real-world tests, GLM-5.3 achieved a remarkable 100% success across five crucial categories: coding, data, real-world tasks, security, and tool use, while maintaining a competitive rubric score of 9.3 and a cost of just $0.28 per task. This contrasts sharply with GPT-5.5, which, despite matching its security performance, demands $1.43 per task with a slightly lower effectiveness in real-world scenarios. The significance of GLM-5.3 lies not only in its cost-effectiveness but also in its comprehensive abilities, making it a strong contender for developers seeking reliable AI solutions across diverse tasks. Its median time-to-first-token stands at 16.3 seconds, which, while slower than some competitors, still positions it favorably for many applications. The release of this model could signal a shift in how organizations leverage AI, offering the capacity for complex task handling without incurring prohibitive costs, thus democratizing access to advanced AI technologies.