GPT-6.1 Terra vs Claude Fable 5.1: benchmarks and token cost

This comparison uses GPT-6.1 Terra (Max) and Claude Fable 5.1 (Xhigh, Default Fallback). It reports the captured benchmark data, with no new hands-on test.

Similar captured scores across different rate tiers

Terra’s captured index is 51.0, compared with Fable’s 49.7. Terra’s listed input and output token rates are one tenth of Fable’s. The close aggregate scores and wide rate gap make this a useful pair for checking effective cost on the same workload. Preserve the Max and Xhigh labels and inspect output quality before drawing a migration conclusion.

How to read the table

The intelligence index is an aggregate measurement, not a percentage of correct answers in your application. Input and output prices are separate rates per million tokens. Coding and math indices are unavailable in this capture and must not be read as zero. Compare first-token latency separately from output-token throughput, and use the same model effort settings for a trial. Choose using task quality, total time and the cost of accepted results; provider features and deployment requirements need their own checks.