Author(s): Caspar Bannink – AI Engineer

Originally published on Towards AI.

Opus 5 Was Just Released And Its …

Impressive, it is very impressive and the ‘comeback’ Anthropic needed. However, when I am choosing a model for a coding, the cheap-looking winner is not automatically the cheap winner. Claude Opus 5 cuts Claude Fable 5 from $10/$50 to $5/$25 per million input/output tokens, while Artificial Analysis places Opus at 61 on its Intelligence Index and Fable at 60. That is a serious release.

After introducing the release, the author argues that headline Intelligence Index scores don’t tell the full story: cost per task can diverge sharply from per-token pricing because effort settings, reasoning time, and token burn affect the final “bill.” Using Artificial Analysis, they compare Opus 5, Fable 5, and GPT-5.6 Sol, emphasizing that Opus’s price cut narrows—rather than eliminates—the gap to Fable, while Sol’s task-cost may be dramatically lower but comes with practical caveats like long-context surcharges. They further caution against simplified claims such as “Opus beat Sol at coding,” noting differences in benchmarks, index versions, fallback/refusal handling, and execution setup. The piece concludes with the author’s routing rule: choose Opus 5 for strong high-end default capability at a lower tier price, choose Sol when completed-task cost is the constraint and context is controlled, and keep Fable 5 when its top-tier behavior is specifically needed—then measure real job outcomes rather than relying on model-name or narrative alone.

Read the full blog for free on Medium.

Join thousands of data leaders on the AI newsletter. Join over 80,000 subscribers and keep up to date with the latest developments in AI. From research to projects and ideas. If you are building an AI startup, an AI-related product, or a service, we invite you to consider becoming a sponsor.

Published via Towards AI