GPT-5.6 Luna Max Underperforms as a Codex Subagent, Dev Finds
daniel_mac8 · x · 2026-08-03
A developer testing various model combinations for an open-source Codex plugin named sol-advisor reported that GPT-5.6 Luna Max performs poorly as a subagent. He speculates this is likely because the model hasn't been post-trained specifically for subagent architectures.
Instead, a combination of GPT-5.6 Sol High (orchestrator) + Sol/Terra (implementers) yielded much better results. The author noted that while Luna Max offers great efficiency and base capabilities, its incompatibility with subagent workflows forced its removal from this specific project.
More from coding & agent
- Dev Workflow: GPT for Coding, Claude for Code Review — mobileraj · 2026-08-03
- Open Source 80 Expert Agent Skills: Clone Famous Thinkers into Coding Agents — tom_doerr · 2026-08-03
- Gradian: Open-Source Tool to Pinpoint Poison Data Causing LLM Fine-Tuning Failures — vylara-ai · 2026-08-03
- Echoverse: Training Agents in Deep Environments Boosts 9B Model Success Rate to 67% — burny_tech · 2026-08-03
- Architecting a Yoga Studio Chatbot: Multi-Model Routing and Low-Cost Memory — omi0009 · 2026-08-03
- Autoexp: A Local-First AI-Assisted Experimentation Workspace — Southern-Whereas3911 · 2026-08-03