{
  "id": 12863482,
  "title": "Claude Sonnet 5.5 vs. Opus 5.5: 42% cheaper and perfect on every run",
  "url": "https://urgent.news/2026/10/08/claude-sonnet-5-5-vs-opus-5-5-42-cheaper-and-perfect-on-every-run",
  "topic": "ai",
  "section": "AI",
  "published": "2026-10-08T12:00:00.000Z",
  "source": {
    "name": "The New Stack",
    "slug": "the-new-stack",
    "url": "https://thenewstack.io/claude-sonnet-5-5-vs-opus-5-5/"
  },
  "original_language": "en",
  "account": "Anthropic introduced Sonnet 5.5 six days after launching Opus 5.5. The new model scored 70.6% on Terminal-Bench 4.0, outpacing Opus 5.5's 66.4% at xhigh effort. Sonnet 5.5 also demonstrated faster output generation and lower token usage. The model was priced at $2 per million input tokens and $10 per million output tokens, half of Opus 5.5's respective costs. However, if a model requires more tokens to complete a task, the savings diminish.\n\nIndependent testing from Artificial Analysis revealed that Sonnet 5.5's per-token savings were not as significant as initially thought. When subjected to max effort, Sonnet 5.5 incurred a cost of $7.67 per task, compared to Opus 5.5's $5.98. Despite the lower price per token, Opus 5.5 proved to be the more budget-friendly option.\n\nDuring rigorous testing, Sonnet 5.5 consistently outperformed Opus 5.5 in all three tests. In the agentic bug fix test, Sonnet 5.5 completed all tasks efficiently and passed all 12 hidden tests. Opus 5.5, on the other hand, finished about 35% faster but came out more expensive overall. In the resolver spec test, Sonnet 5.5 outperformed Opus 5.5 by a 42% margin in terms of cost, while also finishing a little faster. Lastly, in the concurrency bugs test, Sonnet 5.5 emerged as the clear winner, completing all tasks successfully and outperforming Opus 5.5 in both speed and cost.",
  "summary": "Anthropic launched Sonnet 5.5 six days after Opus 5.5. The company says it scores 70.6% on Terminal-Bench 4.0, ahead of The post Claude Sonnet 5.5 vs. Opus 5.5: 42% cheaper and perfect on every run 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."
}