OpenAI has updated its multi-agent framework so GPT-5.6 Sol and Terra can now delegate tasks directly to Luna under Multi-Agent v2, removing a version mismatch that had prevented the speed-optimized model from working smoothly as a subagent in the same workflow. The change addresses a practical limitation for developers building orchestrated AI systems, where Sol is used for heavier reasoning, Terra for a middle ground between capability and cost, and Luna for high-volume tasks where low latency and affordability matter most. With the compatibility barrier removed, a primary agent on Sol or Terra can now call Luna for narrower jobs such as quick lookups or formatting while retaining higher-tier models for more complex decisions. The update makes OpenAI's recommended tiered architecture more workable inside a single agent system, though developers have also flagged that some tools, including Codex, may still need additional configuration changes to take full advantage of Multi-Agent v2.