Anthropic has unveiled Claude Opus 5, which the company framed not as "our best model" but as "Fable 5's performance at half the price." The announced pricing is presented as $5/$25 per million tokens (the source did not elaborate further on the two price points).
What the product promises
According to the release, Claude Opus 5 matches Fable 5 to within 0.5% on certain coding benchmarks while costing roughly half as much. The model is positioned primarily as an affordability improvement rather than a capability leap: Opus 5 aims to offer much of the flagship's performance at substantially lower cost.
Market context
The timing follows Moonshot AI's recent release of Kimi K3, described in the source as the most powerful open‑weight model in the world and made available for free download. When a high‑capacity open model is freely accessible, charging a premium for a closed commercial model becomes harder to justify. Anthropic's response, as presented, was therefore to take its flagship performance and offer it at about half the price.
Why this matters
The announcement is essentially about price competition: in reaction to the cheap, powerful open models coming out of China, Anthropic moved to make its offering more cost‑competitive. The positioning suggests that buyers—developers and enterprises—may now weigh price more heavily than marginal performance differences in choosing models.
Summary
Claude Opus 5 is not presented as a breakthrough in capability but as a strategic pricing move: Anthropic keeps performance close to Fable 5 while markedly lowering the cost to respond to the challenge posed by the free Kimi K3.
