Text models

Fable 5.2 vs. Gemini 4 Carbon for project work

Compare quality and total cost on the same long-running assignment.

Use the same project tasks

Give each model a code change, a review, and a longer task with a checkpoint. Preserve the same inputs, allowed tools, and success criteria.

Measure actual usage

Compare quotes and recorded charges separately. Include elapsed time, failed attempts, and human correction.

Choose per workload

Prefer the model that meets the acceptance criteria for that workload at a lower total cost.

Models in this guide

Related guides and tools

Sources