{
  "id": 1822965,
  "title": "GLM-5.3 Artificial Analysis Benchmarks",
  "url": "https://urgent.news/2026/08/18/glm-5-3-artificial-analysis-benchmarks",
  "topic": "tech",
  "section": "Tech",
  "published": "2026-08-18T22:06:10.000Z",
  "source": {
    "name": "Hacker News",
    "slug": "hacker-news",
    "url": "https://artificialanalysis.ai/models/glm-5-3"
  },
  "original_language": "en",
  "account": "GLM-5.3 (max) is an artificial intelligence model that ranks among the top performers in its intelligence category, offering competitive pricing relative to peers. This model supports text input and generates text outputs, with a substantial context window of 1 million tokens. It excels with a score of 60 on the Artificial Analysis Intelligence Index, significantly outperforming the median score of 35. The model's output verbosity is also notable, producing 170 million tokens during evaluation, compared to the median of 72 million tokens. Pricing for GLM-5.3 (max) is relatively affordable, at $1.40 per million input tokens and $4.40 per million output tokens — both better than the median prices of $1.75 and $10.00, respectively. Evaluating the model on the Intelligence Index cost $1,238.50. This benchmark, which includes a diverse set of tests such as GDPval-AA v2, τ³-Banking, Terminal-Bench v2.1, SciCode, Humanity's Last Exam, GPQA Diamond, CritPt, AA-Omniscience, and AA-LCR, provides a comprehensive evaluation. The model's weights are commercial use restricted, meaning commercial use typically requires a paid license. GLM-5.3 (max) demonstrates strong reasoning capabilities, using chain-of-thought reasoning to tackle complex problems before delivering an answer. However, it does not support image input and is not multimodal, as it only processes text. The model was released on August 18, 2026.",
  "summary": null,
  "key_points": [],
  "editors_take": null,
  "illustration": null,
  "coverage": {
    "outlets": 1,
    "also_reported_by": []
  },
  "ai_generated": true,
  "disclaimer": "Summaries, key points and the editor’s take are written by software from other outlets’ reporting and may contain errors — always check the linked original."
}