Guilherand-Granges, France – March 06, 2025. Anthropic: AI safety company developing Claude language models. Founded by ex-OpenAI members, focusing on safe AI systems.
Credit: PhotoGranary02 via Shutterstock
Anthropic has released Claude Fable 5.1 and Mythos 5.1, the same model with different safeguards, cutting cache read prices by 75% and reporting 52.6% on a scientific research benchmark against 24.7% for its predecessor. Its outputs carry a watermark required by the EU AI Act, and the detection API is in private preview for eligible organisations.
Anthropic has released two new Claude models. Fable 5.1 and Mythos 5.1 are the same model with different safeguards, the company said.
Only one of them is generally available. Mythos goes to vetted users through two access programmes covering cybersecurity and the life sciences. Anthropic says the biology programme was built with the US government.
The developer headline is price. Cache reads fall 75% to $0.25 per million tokens, cutting typical workload costs by about 25% and agentic ones by up to 45%.
The benchmark jump is large on Anthropic’s own numbers. Fable 5.1 scores 52.6% on Terminal-Bench-Science 0.1 against 24.7% for Fable 5, and 55.8% on Terminal-Bench 4.0.
The European part is near the bottom of the post. Anthropic signed the EU’s Code of Practice on transparency of AI-generated content in July, alongside 190 other signatories.
That is why Claude’s output carries a watermark. It applies to models released after 2 August 2026, is invisible without the detection tool, and Anthropic says it carries nothing about the user.
TNW reported last month that Anthropic would ship a free detection API so anyone could check a piece of text.
What has shipped is narrower than that. It is in private preview for eligible organisations, including regulators, law enforcement, media, fact-checkers, independent researchers, educational bodies and EU civil society groups.
The underlying duty is wide. Article 50 requires providers to mark synthetic output in a machine-readable format and make it detectable, and it has applied since 2 August 2026.
Enterprises with their own compliance obligations can apply too. Anthropic says it plans to expand access over time, without giving a date.
The second European gap is the other model. Mythos’s access programmes were built with the US government and are open only to a set of US organisations, two years after Claude arrived in the EU.
Anthropic says it is working to include international partners. European cyberdefenders and life scientists are not in the programme today.
Which is the shape of the launch. The mark exists because of European law, the tool to verify it is in preview, and TNW has reported that removal tools already exist.
Get the TNW newsletter
Get the most important tech news in your inbox each week.
