Anthropic 发布 Claude Fable 5.1 和 Mythos 5.1

MACHINES Anthropic launches Claude Fable 5.1 and Mythos 5.1 by Colin Ryan 2 Sep 2026 Save article Image: © Ralf/Stock.adobe.com Fable 5.1 is now generally available, while Mythos 5.1 is only available through Anthropic’s ‘trusted access’ programmes. Anthropic has launched the latest iterations of Claude Fable and Claude Mythos – “the world’s most advanced models for coding and knowledge work”, according to the company. The AI giant said in a press release that Fable 5.1 and Mythos 5.1 are “the same model, but with different levels of safeguards”. Fable 5.1 is now generally available, while Mythos 5.1 is only available through Anthropic’s “trusted access” programmes – Mythos’s safeguards are specifically designed to support work in cybersecurity and the life sciences, according to the release. The company said that Fable 5.1 is capable of much higher performance than its predecessor and that even when set to medium or low effort, it can achieve similar or better results than Fable 5. The new model can also be used to discover software vulnerabilities, though it can’t exploit them, according to Anthropic. The biggest performance difference between Fable 5.1 and Fable 5 appears to be in relation to agentic scientific research, with the new model capable of more than double the capability in comparison to its predecessor – a jump from a score of 24.7pc to 52.6pc. In terms of price, Fable 5.1 will cost an estimated 25pc less than Fable 5 for “typical workloads, wherever usage is billed by token”. For “highly agentic work”, customers will be charged up to 45pc less. Anthropic is also deploying its new enterprise frontier safeguards (EFS) system, which stores data in cloud infrastructure “controlled entirely” by the customer, not Anthropic. EFS gives customers “complete privacy”, which Anthropic said equates to a zero data retention policy, and will be made available to enterprise customers in phases, beginning later this autumn. Anthropic said that eligible customers will be able to use Fable 5.1 with zero data retention before EFS is available. Mythos 5.1, according to Anthropic, is identical to Fable 5.1 but with “more permissive safeguards” for vetted individuals and organisations whose work is affected by cybersecurity and life sciences restrictions. Currently, Mythos 5.1 is only available to vetted cybersecurity and life science professionals at select US organisations, though Anthropic said it’s coordinating with the US government to expand access to a broader set of domestic and international partners “as quickly as possible”. Mythos 5.1 is now also being used to power Claude Security – the company’s product for scanning codebases for vulnerabilities and suggesting patches for human review. OpenAI readying Astra The debut of Fable 5.1 and Mythos 5.1 comes a few months after their predecessors – particularly Mythos – caused a stir with claims of unprecedented AI capabilities in relation to cybersecurity, including exploitation of vulnerabilities. Security concerns even led to the US government ordering Anthropic to suspend the models for a few weeks. Concern around Anthropic products’ cyber capabilities were reignited recently after the company reported incidents of Claude models (including Mythos) instigating cyberattacks after breaking free during tests. The buzz caused by Mythos in particular ignited further competition with Anthropic’s chief US AI rival OpenAI. According to an announcement earlier this week, the ChatGPT creator is preparing to release its new Astra AI model, which OpenAI claimed can find previously unknown security flaws and develop ways to exploit them across many well-protected systems without a person guiding each step. OpenAI said in a recent blogpost that it has delayed parts of Astra’s development and release over the past few weeks to strengthen and test protections against “cyber misuse and unauthorised model actions”. Astra’s launch is expected “soon”, though access to “its most advanced cybersecurity capabilities” will be more limited, according to the company. “We will continue to test these systems, share what we learn and be clear about what remains uncertain,” said OpenAI. “The models that follow Astra will demand more of us. We will take the time and do the work needed to meet that responsibility.” Don’t miss out on the knowledge you need to succeed. Sign up for the Daily Brief, Silicon Republic’s digest of need-to-know sci-tech news. Related: cybersecurity, Anthropic, research, OpenAI, AI Colin Ryan is a senior reporter/production editor at Silicon Republic editorial@siliconrepublic.com You May Also Like You May Also Like Machines Household names co-sign open letter on AI cybersecurity threat 28 Aug 2026 Machines OpenAI launches GPT-6 Astra 4 Sep 2026 Machines After OpenAI incident, Anthropic finds Claude hacked organisations 31 Jul 2026 Machines Nvidia forms industry alliance after Hugging Face cyber incident 28 Jul 2026 Machines OpenAI launching security AI initiative to compete with Claude Mythos 12 May 2026 Business Reports: Anthropic aims above SpaceX in soon-to-come IPO 21 Aug 2026 More from Science Latest News More More