[AINews] Claude Opus 5: Fable-level performance at Opus price (half Fable)
Latent Space ● Covered by 17 sources
Anthropic dropped Claude Opus 5 on a Friday, and it's landing just a hair below Fable 5 on Epoch's official capability index. But coders testing it say it feels way better than the numbers suggest.
Based on reporting by Latent Space — 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 didn't wait for a normal launch day. Opus 5 showed up on a Friday, and the reception has been a weird mix of benchmark nitpicking and genuine enthusiasm from people actually using the thing to write code and drive browsers.
The headline number comes from Epoch, which measured Opus 5 at an ECI of 159 against Fable 5's 161 — technically a loss, though a thin one. On the software-engineering-specific SWE-ECI, the two models tie at 161. That gap between overall score and coding score is the whole story here: Opus 5 looks like a coding specialist wearing a general-purpose model's benchmark sheet. Anthropic's own messaging leans into
My take — AI-written commentary, not fact-checked reporting
, the company says Opus 5
Read more about this at: Latent Space
