Choosing Gen-AI for automation work

Description and use

Choosing a Gen‑AI model for automation: GPT‑6‑Astra delivered higher quality, fewer steps, lower token usage, and faster execution than Claude‑Fable‑5.1.

Problem

Selecting the most efficient Gen-AI model for a workflow automation project at a growing agency.

Outcome

GPT‑6‑Astra produced higher quality results in four steps, used 1.5x fewer tokens, and executed twice as fast as Claude‑Fable‑5.1, leading to its selection for future planning tasks.

Workflow

• 1. Define the complex instruction to be automated.
• 2. Run the instruction through GPT‑6‑Astra and record steps, tokens, and time.
• 3. Run the same instruction through Claude‑Fable‑5.1 and record metrics.
• 4. Compare quality, steps, token usage, and execution time.
• 5. Select the model that meets quality and efficiency criteria (GPT‑6‑Astra).

Evidence

GPT‑6‑Astra was able to achieve excellent results in a shorter number of steps – specifically, a complex instruction required four steps to execute. In comparison, Fable‑5.1 took five steps and produced inferior results, consuming 1.5x more tokens and taking twice as long.

Source