Picking models for orchestrator and subagent roles in Codex
Melvin Vivas · X post · 2026-09-26 · Open on X
Topics: AI Agents, Tool Use & MCP, LLM Fundamentals, AI Dev Tools & Productivity · Level: intermediate
Summary
After testing GPT-6 Sol, the creator is staying with a Codex multi-agent setup: Astra as the orchestrator and Luna as subagents. His reason is that Sol costs too much to run as a subagent and is not smart enough to be the orchestrator. The lesson is to choose a model for each role based on cost and intelligence.
Key points
- Multi-agent setup in Codex: one orchestrator model plans and delegates the work, and subagent models carry out the tasks.
- The creator's preferred setup is Astra as orchestrator and Luna as subagents.
- GPT-6 Sol was rejected for both roles: too expensive for a subagent and less intelligent than Astra as an orchestrator.
- Subagents run many times, so cost matters most there. The orchestrator makes the key decisions, so reasoning quality matters most there.
Resources mentioned
- OpenAI Codex · tool · openai.com · paid · recommended by both Bashiri Smith & Melvin Vivas
OpenAI's coding agent. In the diagram it writes code, fixes review findings and drives the build loop. The creator also used it to make this video.
Also in: Developers moving from ChatGPT/Codex to Claude (Melvin Vivas on X · notes), Creator Opinion: Codex Is Still the Best at Computer Use (Melvin Vivas on X · notes), OpenAI Codex Ambassador applications are open (Melvin Vivas on X · notes), GPT-Live-1.1: Multilingual Voice Update for ChatGPT and Codex (Melvin Vivas on X · notes) and 264 more - Luna · tool · openai.com · paid
The other model/agent the classifier routes tasks to; the post doesn't describe it further.
Also in: GPT-6.1 Sol May Beat Luna for Subagents (Melvin Vivas on X · notes), GPT-6 Sol Ultra Subagents Use Up Limits Fast (Melvin Vivas on X · notes), GPT-6 Sol vs Opus 5.5 in a Livestream Comparison (Melvin Vivas on X · notes), DigitalOcean Serverless Inference Now Serves OpenAI GPT-6 Models (Melvin Vivas on X · notes) and 14 more - GPT-6 Astra · tool · openai.com · paid
The model announced in the quoted launch post, pitched as the developer's most capable model for work, coding, science and cybersecurity, and able to operate a computer.
Also in: Glimpse: A Visual Reasoning Model That Uses a Visual Scratch Space (Melvin Vivas on X · notes), Is AI Getting Cheaper? How to Measure LLM Cost Per Task and Per Intelligence Level (Melvin Vivas on X · notes), GitHub Copilot Adds Intelligent Local Model Routing (Project HydraFusion) (Melvin Vivas on X · notes), Comparing Opus 5.5, GPT-6 Astra and Grok 4.7 on OpenRouter (Melvin Vivas on X · notes) and 55 more
Try this
- When building a multi-agent setup, compare models separately for the orchestrator and subagent roles, weighing cost against intelligence.
More in AI Agents, Tool Use & MCP
- Grok Team Bots: Shared AI Teammates in Slack and Grok
- AI Agents vs AI Workflows: How to Choose for Production
- Grok Bot Templates: Sharing, Publishing and Safely Installing Bots
- Open-source Codex skill: Astra/Sol orchestrator with Luna subagents
- Cost-Efficient Codex Subagents: Astra/Sol Orchestrator + Luna Workers
- Google's multi-agent framework for consistent long-form video