Anthropic Releases Claude Opus 5.5: Fable 5.1-Level Performance at 40% Lower Running Cost Than Opus 5
MarkTechPost Asif Razzaq ● Covered by 3 sources
Anthropic launched Claude Opus 5.5, saying it matches Fable 5.1 on most work. It also runs 40% cheaper than Opus 5, but you can only use it through Anthropic’s API.
Based on reporting by MarkTechPost, Asif Razzaq — read the original for the full story.
Summary, retelling and take written by AI under human oversight; images are AI-generated illustrations. How we work · Report an error
Anthropic has opened up Claude Opus 5.5, the first model in its new Claude 5.5 line, and it’s pitching the release as a serious step forward on cost as much as capability. The company says the model performs at the level of Claude Fable 5.1 on most work, while cutting typical running costs by 40% versus Opus 5 at default settings.
The headline benchmarks are strong, though not perfectly tidy. Anthropic says Opus 5.5 leads its own tests in agentic coding, computer use, and knowledge work, and it posts the top score to date on the company’s automated behavioral audit. But the comparison gets messier once the numbers spread out: GPT-6 Astra still leads on Terminal-Bench-Science and AutomationBench, and Anthropic itself says benchmark margins are becoming a less reliable guide.
What seems more important is the price-performance shift. On FrontierCode, Opus 5.5 at default medium effort scores 54.6%, ahead of GPT-6 Astra’s best 53.3%, while costing about a fifth as much per task. On CursorBench, it reaches 52.5%, which Anthropic says is 11 points above GPT-5.6 Sol’s best score at around a third of the cost. The model also uses fewer tokens per task, cache reads drop 60%, and output generation is more than 30% faster than Opus 5.
Early testers seem to have pushed those savings hard. Anthropic cites a 680,000-line code migration finished in under a day, a 200,000-line codebase audit and fix completed in under 3 hours, and an internal HAProxy port from C to Rust that finished in 9.5 hours. In that same HAProxy test, Fable 5.1 took 12 hours, and Opus 5.5 cost 51% less.
The catch is that this is still a closed model. There are no released weights, so self-hosting is off the table, and access runs through the Claude Platform plus Amazon Web Services, Google Cloud, and Microsoft Azure. Anthropic is also keeping the guardrails tight: thinking can’t be turned off, output is watermarked for EU AI Act compliance, and cybersecurity and biology features are wrapped in the same kind of controls the company used for Fable 5.1.
My take — AI-written commentary, not fact-checked reporting
Anthropic is doing the boring thing that actually matters: making a frontier model cheaper instead of just louder. That will annoy the open-model crowd, but the market keeps rewarding closed systems that ship useful controls, usage limits, and cleaner economics. The real story here is not a benchmark trophy; it’s that efficiency is becoming the new flex.
Read more about this at: MarkTechPost
