Anthropic sells two AI models built on the same weights at the same price, but only one can be bought freely. The other requires an application and a review.
Anthropic launched two AI models on 1 September 2026 that are the same underneath and cost the same. Anyone can buy Claude Fable 5.1. Claude Mythos 5.1, which has fewer safety blocks and performs better on agent tasks, is only for organizations and people that pass a review.
Fable 5.1 replaced Claude Fable 5. It is available on the Claude API, Amazon Bedrock, Google Cloud and Microsoft Foundry. Mythos 5.1 shipped the same day on the same weights, with the cyber and biology safeguards tuned down for vetted defenders. Safeguards are the filters that make an AI refuse risky requests. The earlier pair, Fable 5 and Mythos 5, were suspended worldwide two days after their 9 June 2026 launch by a US Commerce Department export-control directive. They were later restored for a limited set of vetted organizations.
Anthropic tested the access levels using a different model, Claude Opus 5.5, on ten multi-step cyber challenges with five attempts each. With no program access, every task was blocked at the first prompt. Under Defense Access, 46 of 50 trials were blocked somewhere along the way. Under Red Team Access, 34 of 50 were completed, close to the 67.6% success rate the model reaches with no safeguards at all. The model is the same in every case. What changes is how often it gets stopped.
Security teams doing defensive work, incident response, malware analysis or authorized red-teaming (hired, permitted attack testing) are the audience for Mythos. Defense Access is open to individuals on a paid plan, so one researcher with a track record can apply. The review takes longer than the setup, so applying early matters. Everyone else is meant to use Fable 5.1.
Comparing scores is tricky. Anthropic pairs Fable 5.1 only with Fable 5, so there is no direct table. Mythos 5's 88.0 on Terminal-Bench 2.1 cannot be compared with the 60.9 on Terminal-Bench 4.0, because they are different tests. Artificial Analysis scores Fable 5.1 at 53, and Opus 5.5 at 58.
If you do defensive security work, apply for Mythos access early, because the review takes longer than the setup.
Lawmakers in both parties want Nvidia's CEO to answer questions about a reported tie to an alleged Chinese intelligence asset, and chip rules could shift as a result.
Read more: Lawmakers from both parties want Nvidia's CEO to answer questionsDeepSeek says 1.54 million people use its AI agent platform each day. If you run a business, cheaper AI helpers may soon be offered to you.
Read more: DeepSeek says its AI helper platform now has 1.54 million daily usersOpenAI has about $11 billion in hand at a $1.25 trillion value, below its first ask. Microsoft and NVIDIA shareholders have reason to watch what comes next.
Read more: OpenAI raises $11 billion, but at a lower price than it wanted
