Model Routing for Coding: Estimate Task Difficulty, Then Pick a Model

Melvin Vivas · X post · 2026-09-16 · Open on X

Topics: LLM Fundamentals, AI System Design & Architecture, LLMOps, Deployment & Monitoring · Level: intermediate

Summary

Melvin Vivas suggests using the newly announced Jev model for coding by first estimating how hard a task is and then sending it to a suitable model. The quoted launch post says Jev was trained with a new method called RLCD and claims it is 20-200x faster. The lesson is the model-routing pattern: cheap or fast models handle easy tasks and stronger models handle hard ones.

Key points

• Routing pattern: estimate a task's difficulty first, then send it to the model that fits.
• Proposed use of Jev: act as a fast classifier or router in front of coding models.
• The quoted launch post says Jev is trained with a new method called RLCD and claims it is 20-200x faster.
• The launch post is from a self-described co-inventor of ChatGPT who spent 2 years in stealth.
• Routing can lower cost and latency because the strongest model is only used where it's needed.

Resources mentioned

• Jev · tool · typesafe.ai · paid
  A decision model that makes turn-level forecasts on AI agent conversations using only structural signals (turns, tool calls, workflow stages, timing).
  Also in: Generating a Repo Promo Video with a Claude Skill on Sonnet 5.5 vs Opus 5.5 (Melvin Vivas on X · notes), GLiDE by Fastino Labs: A Post-Trainable Reasoning Decision Model (Melvin Vivas on X · notes), Using Jev as a Reranker to Augment RAG Retrieval (Melvin Vivas on X · notes), Using Jev (TypeSafe AI) as a Decision Model for Email Classification (Melvin Vivas on X · notes) and 12 more

Try this

• Try putting a difficulty-estimation step in front of your coding agent and routing easy and hard tasks to different models.
• Build a coding-task router that scores how hard a task is and sends it to a cheap, fast model or a frontier model.