{
  "id": 18439658,
  "title": "Closed AI models dominate the price-performance frontier",
  "url": "https://urgent.news/2026/11/06/closed-ai-models-dominate-the-price-performance-frontier",
  "topic": "ai",
  "section": "AI",
  "published": "2026-11-06T02:49:56.000Z",
  "source": {
    "name": "Hacker News",
    "slug": "hacker-news",
    "url": "https://village.dev/blog/closed-ai-models-now-dominate-the-price-performance-frontier/"
  },
  "original_language": "en",
  "account": "The latest AI model releases are increasingly shifting the price-performance frontier in favor of closed models, according to a new study using Village's LLM Atlas. Open models often lead in value during bursts, with a strong, inexpensive release from DeepSeek capturing a wide range of budgets until a closed model surpasses it in capability. Recent releases from closed models have been both more affordable and more powerful than their open counterparts, narrowing the price range where open models excel. The study estimates each model's capability score from public benchmark results and cost per task using a separate cost model, with costs relative to GPT-5 at high effort. For each date, the researchers identify the most capable model within 201 evenly spaced budgets spanning from 1% to three times GPT-5's cost. The scores are based on data as of October 31, 2026, and are updated as new results become available. The cost figures reflect what evaluators paid through hosted APIs, while open models may be cheaper to run on personal hardware. Using models scored on at least 10 benchmark families or focusing the budget range to 3% to one times GPT-5's cost yields the same pattern. However, expanding the budget range to 0.3% to ten times GPT-5's cost increases the share of the cheapest end occupied by tiny open models to 16%.",
  "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."
}