Anthropic has just presented Claude Narrative 5 and Claude Mythos 5the most advanced artificial intelligence models that the company has put on the table to date.
From Anthropic they are quite direct about it. The company assures that Narrative 5 surpasses in capabilities any neatly-liked model they have made available so farplacing it at the state of the art in virtually every AI capability benchmark they have measured, including utility engineering, knowledge work, synthetic vision, scientific research, and much more. The promise is clear: the longer and more complex the task, the greater the advantage of Narrative 5 over previous models from the same company.
What makes all of this especially interesting is that both Narrative 5 and Mythos 5 share the same underlying model. The difference between the two is not in the engine, but in the safeguards. Claude Narrative 5 is the safe version for neatly-liked usewhile Claude Mythos 5 is the same sinful but with certain restrictions liftedinitially intended for a small group of cyber defenders and critical infrastructures through Project Glasswing, an initiative in collaboration with the United States government.
Why Claude Narrative 5 is a game changer
The first thing that stands out when reviewing the technical data is the level of performance that Anthropic has documented with real users. Stripe, for example, reported that Narrative 5 compressed months of engineering into a matter of days. In a 50-million-line Ruby code pipeline, the model completed a migration at scale in one day that would have required weeks of work for an entire team. That’s not a marginal improvement, that’s a paradigm shift.
In the financial field, the results are equally compelling. IMC confirmed that Narrative 5 passed almost all of its trading analysis assessments, including fact finding, conceptual reasoning, and expected value analysis. Furthermore, in the Hebbia finance benchmark for senior level reasoning, Narrative 5 scored the highest of any model availablewith notable gains in document-based reasoning and graph interpretation.
The vision capabilities also deserve separate mention. The model can reconstruct the source code of a web application from screenshots and extract precise numbers from detailed scientific figures. For those who remember the limitations of previous versions, this represents a leap forward in actual fact.
Claude Mythos 5, the model reserved for the most critical
While Narrative 5 is available to any user today, Mythos 5 operates on a different level. Anthropic characterizes it as the model with the most powerful cybersecurity capabilities of any AI system in the world. It is not a light comment. The company’s own researchers subjected the model to tests where it demonstrated an indispensable ability to discover and exploit utility vulnerabilities, surpassing even the capabilities of expert human operators.
That is why access to Mythos 5 is currently restricted. Project Glasswing, developed in direct collaboration with the US government, channels this model towards cyber defenders and critical infrastructure providers who need tools of this caliber. In life sciences research, Anthropic’s internal experts reported a Up to ten-fold acceleration in aspects of the drug design processwith the model autonomously executing tasks that a scientist normally completes, from selecting binding sites to recovering from process failures.
Another striking fact is that Mythos 5 is the first Anthropic model that it invented consistently. novel and compelling scientific hypotheses. In blind comparisons against Opus-class models, company scientists preferred Mythos’ molecular biology hypotheses in about 80% of cases, and several of those approaches have already advanced to experimental evaluation.
Availability and price: more accessible than expected
One of the most surprising details of the launch has to do with the price. Narrative 5 and Mythos 5 are available at $10 per million tokens in and $50 per million tokens outwhich is less than half the cost of Claude Mythos Preview. For developers who have been working with that model, it is very welcome news.
For users of subscription plans, Narrative 5 is included at no additional cost on Professional, Max, Group, and Endeavor plans through June 22. After that date, its use will require consumer credits, although Anthropic has made clear that its intention is to reinstate it into standard subscription plans as soon as capacity allows.
Regarding safeguards, the model incorporates advanced security classifiers that detect potentially dangerous uses related to cybersecurity, biology, chemistry and model distillation. When a query triggers those classifiers, the response is automatically generated by Claude Opus 4.8, a highly capable model in its own right. Anthropic notes that this occurs in less than 5% of sessions, so for the vast majority of users the performance is equivalent to that of pure Mythos 5.
The panorama that characterizes this launch is that of a company that has decided to no longer wait to bring its most powerful models to the general public, betting on a balance between security and accessibility that, until now, few in the industry had managed to articulate with such clarity.
Keep reading:
• ChatGPT vs Gemini vs Claude: which has the cheapest concept top rate and what you get by paying for it
• Anthropic created an AI so dangerous that it cannot release it to the public: thus Project Glasswing was born
• They manage to breach Apple’s M5 chip: Is your MacBook in danger?
