Cheap tricks for AI

2026-07-07 · Source: Matthew Berman · Field: Technology & Digital — Artificial Intelligence & Machine Learning · Depth: Intermediate, quick

Summary

Model routing offers a straightforward strategy to reduce AI operational costs by over 90 percent. This technique involves leveraging a high-capability "frontier model," such as Fable, for initial planning tasks, like generating a feature specification. This detailed plan is then handed off to a significantly more economical "coding model" for the actual execution of the task. After the coding model completes its work, the output is routed back to the premium frontier model for a final review or "once over." This sequential delegation of tasks, from high-cost planning to low-cost execution and back for final verification, enables substantial cost savings while maintaining quality.

Key takeaway

For AI Engineers optimizing operational budgets, implementing model routing can drastically cut costs. If you are currently relying solely on expensive frontier models for end-to-end tasks, consider segmenting your workflow. You should use premium models for strategic planning and final verification, while delegating execution to more affordable, capable coding models. This approach can yield over 90 percent savings on your AI bill without compromising output quality.

Key insights

Model routing delegates high-level planning to expensive AI and execution to cheaper models, saving over 90%.

Principles

• Delegate complex planning to premium AI.
• Execute routine tasks with cost-effective models.
• Re-verify outputs with high-tier AI.

Method

Use a frontier model for planning (e.g., spec writing). Pass the spec to a cheaper coding model for execution. Return output to the frontier model for final review.

In practice

• Implement a two-stage AI workflow.
• Integrate Fable for initial planning.
• Utilize cheaper coding models for execution.

Topics

• Model Routing
• AI Cost Optimization
• Frontier Models
• Coding Models
• AI Workflow
• Fable

Best for: Director of AI/ML, AI Engineer

Related on AIssential

• How Small AI Models Can Cut Your AI Costs By 10x — Over-reliance on large frontier models for all AI tasks leads to unnecessary costs and latency.
• The Five Kinds of Model Routers That Cut AI Costs — Model routers dynamically select optimal AI models for tasks, significantly reducing operational costs without sacrificing quality.
• Cut your AI cost IN HALF (EASY) — Model routing optimizes AI costs by matching task complexity to model capability, using expensive models for planning and cheaper ones for execution.
• I Replaced One $15-Per-Million-Token Frontier Model With Five Cheap Ones Arguing With Each Other. — Ensembling multiple cheap models can outperform a single expensive frontier model in production.
• ACRouter picks the smartest AI model per task, beating Opus-only setups by 2.6x on cost — Dynamic, feedback-driven AI model routing, like ACRouter's C-A-F loop, significantly enhances cost-efficiency and performance over static approaches.
• WORST AI Mistakes by Companies — Model routing significantly reduces AI inference costs by intelligently delegating tasks to appropriate models.

See Counsel's argued verdicts on the open AI decisions leaders are weighing →

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by Matthew Berman.