Anthropic announces Claude Fable 5.1 and Mythos 5.1 models

Anthropic stated that Fable 5.1 would cost an estimated 25% less than Fable 5 for “typical workloads” wherever the usage is billed by token [File] | Photo Credit: REUTERS Anthropic announced the release of its new Claude Fable 5.1 and Claude Mythos 5.1 models on September 1, referring to them as “the world’s most advanced models for coding and knowledge work.”The two models are the same except for their safeguards, according to the Claude-maker, with Fable 5.1 being generally available, while Mythos 5.1 is available only through the company’s trusted access programmes.As for pricing, Anthropic stated that Fable 5.1 would cost an estimated 25% less than Fable 5 for “typical workloads” wherever the usage is billed by token. The company explained this was because it was reducing its pricing on models reading inputs that were already processed and stored.For highly agentic work, the company said users could expect savings of up to approximately 45%.Cybersecurity safeguardsAnthropic reported that Fable 5.1 could be used to discover software vulnerabilities, but not to develop exploits.As for Mythos 5.1, Anthropic reported it was “significantly less likely than Mythos 5 to try to access resources outside of its test environment when assigned an otherwise impossible task.”This comes in light of rising concerns about AI models breaking out of their restricted testing environments in order to seek answers and solutions on the internet, even if this requires hacking into other companies.The AI firm backed by Google and Amazon also tested the model’s ability to create biological or chemical weapons; this is a risk factor that CEO Dario Amodei had cited when discussing the threats that open-weight AI models could enable.According to the company, testing involved red-teaming, automated evaluations, and a tabletop exercise that paired PhD-level biologists with AI experts, to test the models’ capabilities.“Mythos 5.1’s capabilities are greater than those of Mythos 5. However, our evaluations indicate that it still falls short of the next risk tier defined in our Responsible Scaling Policy. We are therefore deploying Mythos 5.1 with the same safeguards that we applied to Mythos 5, which restrict access to research biology capabilities,” confirmed Anthropic.However, the company admitted that the Claude Mythos 5.1 model could still sometimes bypass approvals and auto-mode classifiers.Anthropic has also been vocal in calling out alleged instances of distillation by Chinese AI firms training their models on Claude’s output. The company said that Fable 5.1 had “strengthened mechanisms” to complicate distillation attacks.As part of this, new API accounts will no longer be able to manually edit Claude’s prior context in a multi-turn conversation if they want to preserve the transcript of Claude’s prior thinking.AI watermarksSince Anthropic in July signed the EU AI Act’s Code of Practice on Transparency of AI-Generated Content, the company will add a watermark to the outputs of its models released after August 2, 2026. This also applies to AI-generated text and translations.The Claude-maker assured users this would not degrade the performance of their model, and noted it was rolling out a watermark detection API in private preview for eligible users, such as regulators and the media.However, users have raised concerns about the reputations of original authors being harmed through false positive results. Published - September 02, 2026 02:29 pm IST