{
  "id": 19538669,
  "title": "Anthropic built a model that beats Opus. It recommends Opus anyway.",
  "url": "https://urgent.news/2026/11/11/anthropic-built-a-model-that-beats-opus-it-recommends-opus-anyway",
  "topic": "ai",
  "section": "AI",
  "published": "2026-11-11T19:15:21.000Z",
  "source": {
    "name": "The New Stack",
    "slug": "the-new-stack",
    "url": "https://thenewstack.io/anthropic-fable-opus-cost/"
  },
  "original_language": "en",
  "account": "Anthropic unveiled Fable 5.2, a coding model that bests Opus 5.5 on two key benchmarks and delivers results for half the price. Despite this, Anthropic still suggests that most developers continue using Opus, labeling it as the \"strongest daily driver\" for coding and knowledge work. The reason behind this recommendation is that Fable is primarily designed for large software migrations, multi-agent builds, and extensive research projects. However, the performance data introduces complexity to the recommendation.\n\nFable's higher cost per token ($8 million input tokens and $40 million output tokens) compared to Opus ($4 and $20) doesn't always translate to higher costs. In an internal experiment, 1,000 Fable 5.2 agents formalized Fermat's Last Theorem in 30 hours, while Opus 5.5 only managed 26% completion after 18 hours and showed little progress by the 30-hour mark. This suggests that Fable could be more economical for certain tasks, even with its higher per-token cost.\n\nAnthropic also proposes using Fable as a coordinator for cheaper models. Developers can delegate simpler tasks to Opus 5.5 or Sonnet 5.5, or consult these models for more challenging problems. This multi-model architecture is intended to save costs by utilizing cheaper models for tasks they can handle effectively. However, the context-switching between models can introduce complications and additional costs. For instance, Fable can read Opus's generated \"thinking blocks,\" but Opus cannot understand Fable's outputs.\n\nAnthropic advises developers to initially use Opus and switch to Fable when Opus fails to meet project requirements. This capability-based switching rule might not always lead to the lowest total execution cost, as it doesn't consider the lowest cost scenario. Fable's FrontierCode results indicate it can achieve comparable performance to Opus at roughly half the task cost, suggesting that developers could potentially benefit from switching to Fable earlier than Anthropic suggests.",
  "summary": "Anthropic just released a coding model that beats Opus 5.5 on two major benchmarks and can complete certain difficult tasks The post Anthropic built a model that beats Opus. It recommends Opus anyway. appeared first on The New Stack .",
  "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."
}