What's this about?
Astra costs $10 per million input tokens and $50 per million output — 2.5× Sol's $4/$20, and exactly matching Anthropic's Fable 5.1. A fast mode offers up to 2× speed at 2× price.
What exactly is per-token pricing?
API usage is billed by text volume in tokens (roughly ¾ of a word each), with output costing more than input. Agentic workloads burn tokens in bulk, so a 2.5× multiplier changes what's economical to automate, not just the monthly bill.
Why is this worth noting?
HN read it as the end of OpenAI's undercutting era: "OpenAI finally found an edge to stop selling cheap." Price-matching Fable to the dollar signals the frontier tier now prices on capability, not market share — and subscription users immediately started calculating how fast the new model burns their quotas.