Fable 5.1 and GPT-6 Astra Launch in Same Week, Compete Across Benchmarks

Anthropic's Fable 5.1 and OpenAI's GPT-6 Astra debuted in the same week, setting off direct comparisons across benchmarks and real-world tasks. On formal evaluations, Astra leads Fable 5.1 on FrontierMath Tier 4 (97.6% vs 87.8%), ARC-AGI-3 (62.7% on the official harness), and the Epoch AI ECI (169 vs 163). On Vending-Bench 2, a simulated autonomous business task, Astra averaged $15,515 versus Fable 5.1's $5,422, with Astra's worst run finishing above Fable's best. In a Factorio Space Age experiment, Astra reached seven science packs while Fable 5.1 reached only two. Astra is priced at $10 per million input tokens and $50 per million output tokens, which one item notes is on par with Fable 5.1. Despite Astra's benchmark leads, some users in an informal poll described by Zvi Mowshowitz still prefer Fable 5.1 for agentic coding reliability, and user preference shifted from a roughly 2:1 Claude advantage before either launch to nearly even between the two models.