Fable 5.1: The New Best AI Model Is Once Again More Expensive
Trending Topics Jakob Steinschaden ● Covered by 4 sources
Anthropic’s new Claude Fable 5.1 is now top of the AI rankings. But the benchmark leader also costs more per task than its predecessor.
Based on reporting by Trending Topics, Jakob Steinschaden — 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 launched Claude Fable 5.1, and its near-twin Claude Mythos 5.1, which is only available to vetted organizations. On Artificial Analysis’ measurements, Fable 5.1 now sits at the top of the pack. That’s the headline. The less flattering detail is that the best model is again the pricier one.
Anthropic is cutting cache-read prices by 75 percent, and says that should translate into about 25 percent savings for typical workloads, or roughly 45 percent for heavy agentic use. But Artificial Analysis says the bill still goes up once token usage is counted. A single Intelligence Index task costs $3.76 with Fable 5.1 at max effort, versus $3.14 with Fable 5. That’s about 20 percent more. The reason is simple enough: the model uses a lot more output tokens.
The performance numbers are hard to ignore. Artificial Analysis gives Fable 5.1 a 66 on its Intelligence Index at the “max” setting, which it says is the highest score it has ever recorded. That beats Claude Opus 5 at 63, Claude Fable 5 at 62, and GPT-5.6 Sol and Grok 4.6 at 61. Anthropic also reports gains on several individual tests, including Humanity’s Last Exam, Terminal-Bench v2.1, SciCode, and a banking scenario called τ³. On agentic science tasks, Anthropic says the score rose from 24.7 percent to 52.6 percent.
Still, this isn’t a clean victory lap. The cheapest part of the launch is the cache read cut, and it mostly offsets the model’s higher appetite rather than erasing it. Artificial Analysis says the cache-read discount saves about $1.40 per task. Without it, the same task would land around $5.16. Anthropic is effectively selling more performance and charging for the appetite that comes with it.
Fable 5.1 keeps the same one-million-token context window, supports text and image inputs, and is already available through the Claude API and major cloud platforms. Claude Pro subscribers get a sting in the tail: they pay extra to use it, while Claude Opus 5 remains the best model included in the subscription.
My take — AI-written commentary, not fact-checked reporting
This is the AI market in one tidy paragraph: better scores, fatter bills, and a press release to make the math look polite. The real story isn’t that Anthropic built the best model; it’s that “best” still comes with a usage tax. Anyone shopping for value should probably keep an eye on the model that’s merely close, not the one that hoovers tokens like a shop vac.
Read more about this at: Trending Topics
Related stories
Opus Outshines Even Fable, Inside the Hugging Face Hack, AI Companies Spend Big for Compute
The Batch ·
41