Anthropic launches Claude Fable 5.1 after inking $35B cloud deal with Lambda

Anthropic has a pair of new models out: Claude Fable 5.1 and Claude Mythos 5.1. The company calls them its most capable large language models yet, and the timing is hard to miss. The launch landed a day after Anthropic signed a $35 billion infrastructure deal with Lambda, and a week after another, larger hardware contract with Nscale.

Fable 5.1 and Mythos 5.1 are basically the same model, but Anthropic has split them by permission. Fable 5.1 is the general release. It won’t be used for sensitive cybersecurity and biology work. Mythos 5.1 is the locked-down version for trusted organizations, and Anthropic says it has the “strongest cyber capabilities of any model we’ve released.”

The benchmark claims are aggressive. Anthropic says Fable 5.1 set records across several tests, including a 52.6% score on Terminal-Bench-Science 0.1, which it says measures research-task performance. That was more than double the previous Fable 5. It also scored 13.% higher on Terminal-Bench 4.0 for coding, and did well on several non-technical task collections.

The other big change is cost. Anthropic says an improved prompt caching system makes Fable 5.1 about 25% more cost-efficient on “typical workloads” than the earlier model, with savings of up to 45% for applications that lean heavily on AI agents. That suggests the company has cut some hardware use even as it lines up huge amounts of it. The Lambda deal alone is part of a larger buildout that may amount to somewhere between 350 and 437 megawatts, based on Jensen Huang’s recent estimate of AI infrastructure costs.

That Lambda capacity is tied to a data center under construction in Nueces County, Texas, by Hut 8. Nvidia will hold the lease on the campus, which may help lower risk for the parties involved. Hut 8 expects to connect the site to the grid in the first quarter of 2027, with the first data hall following about six months later.