Claude Haiku 5.5 (Xhigh) vs GPT-6 Astra (Medium)

What the captured data shows

Haiku 5.5 Xhigh has lower listed prices than GPT-6 Astra Medium and a lower intelligence index. Haiku’s reported output throughput is higher, but Astra’s first-token delay is shorter. This pair illustrates why a reasoning setting and a throughput number cannot, on their own, describe the experience of using a model.

Which model fits your workflow

Measure a complete job at these exact settings. Check whether extra reasoning produces a useful improvement and whether waiting before output is acceptable for the interface. The missing individual benchmark scores leave task-specific superiority to be established by your own controlled examples.

Decision takeaway

The table supplies separate signals for cost, aggregate intelligence and timing; no single signal settles the choice for every workload.

These figures come from Franklin’s Artificial Analysis-sourced model records captured on 2026-10-11. Read the individual model entries for the data snapshot: Claude Haiku 5.5 (Xhigh) and GPT-6 Astra (Medium). Missing measurements remain unavailable; this page does not report a new hands-on test.