Anthropic Ships Claude Fable 5.1 with Better Benchmarks, Lower Costs, and Built-In Text Watermarking

Summary

Anthropic released Claude Fable 5.1 and Claude Mythos 5.1 yesterday, calling them “the world’s most advanced models for coding and knowledge work.” Fable 5.1 beats both its predecessor and OpenAI’s GPT-5.6 Sol across major benchmarks, including a score of 52.6 percent on Terminal-Bench-Science 0.1, more than double what Fable 5 achieved on the same test. It costs about 25 percent less than Fable 5 for typical workloads and up to 45 percent less for highly agentic tasks, due to reduced cache-read pricing. The update also bakes invisible watermarking into all Fable 5.1 text, with a detection API rolling out to eligible organizations in private preview. Claude Code users should see roughly 60 percent fewer cybersecurity false positives.

Why it matters

For everyday Claude users, the cost drop is the most tangible change. If you use Claude for writing, research, or any long-running task, Fable 5.1 gets you equal or better quality for less money. The watermarking is worth understanding before you start publishing anything Claude helped you write. Any text Fable 5.1 generates carries a hidden signal identifying it as AI-written, even if you paste it into another document. That is the kind of thing content creators and writers should know upfront.

HTD Says

Anthropic just shipped a Claude model that costs 25 percent less, beats GPT-5.6 Sol on key benchmarks, and quietly embeds an invisible watermark in every piece of text it generates, which is something every writer or publisher using Claude should understand before shipping anything. Watermarking or not, we should all just assume that content is now being generated by AI.

I purposefully stay a bit behind on my AI versions. If I find something that works, I tend to stick with it, although I do ask third-party AIs to compare models for me. Currently, I’m on 5.6 Terra for Codex and Sonnet 5 for Claude. Are they the best choices? Probably not, so I will have to do another assessment.

If Anthropic and others are claiming that Fable 5.1 is better than 5, it might be worth looking at. But I try not to look at the previous model, but rather at ones slightly before that. Gemini, my source, says that for my work, 5.6 terra is still the best choice. Similarly, Gemini recommends sticking with Sonnet 5 as a good all-around workhorse. Guess I’m not changing.

Source:

MacRumors