Match Model and Reasoning Level to Task Complexity in Coworker
Melvin Vivas · X post · 2026-09-05 · Open on X
Topics: AI Dev Tools & Productivity, LLM Fundamentals · Level: intermediate
Summary
While building a headless mode for Coworker, the creator runs GPT-6 Astra on its low reasoning setting so his usage limits last longer. He switches to the cheaper Luna model when a task is simpler. The takeaway is to route each task to a model tier based on how hard it is.
Key points
- The creator is building a headless (non-interactive) mode for Coworker.
- He uses GPT-6 Astra on the 'low' reasoning setting to make usage last longer.
- He deliberately switches to Luna for simpler tasks.
- Matching the model and reasoning effort to the task saves cost and quota.
Resources mentioned
- Coworker (donvito/coworker) · repo · github.com · free
The creator's agent/bot tool. Version 0.7.0 adds a Primary Bot, free web search, an activity dashboard and support for open models through OpenRouter.
Also in: Coworker: a local-first desktop AI coworker that works with any model (Melvin Vivas on X · notes), Liquid AI's LFM2.5-24B-A2B: a local MoE model for agent tasks (Melvin Vivas on X · notes), Coworker: A Desktop AI Assistant App, Not a Coding Agent (Melvin Vivas on X · notes), Coworker v0.7.2: Open-Source Desktop AI Assistant (Melvin Vivas on X · notes) and 59 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 - 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), Picking models for orchestrator and subagent roles in Codex (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) and 14 more
Try this
- Use a low reasoning setting on expensive models to make usage last longer.
- Switch to a cheaper model like Luna for simple tasks.
More in AI Dev Tools & Productivity
- Building and Deploying a Flappy Bird Clone with GPT-6 Astra in Codex
- Inspect Codex Traces with jsonl-viewer
- Codex Agent Command Center: the `codex agents` Command
- Codex Plus Plan: Use Luna and Save Sol and Astra
- Luna, Not Sol, as the Daily Driver on Codex Plus
- Launching Cursor Coding Tasks from Grok Bot