OpenAI has released Multi-Agent v2, a major update to its multi-agent framework that now allows GPT-5.6 Sol and Terra models to delegate tasks to Luna, the speed-optimized variant. This upgrade resolves a version compatibility issue between v1 and v2 that had prevented collaboration between these models for weeks. The new three-tier architecture enables developers to build more flexible workflows, with Sol handling complex reasoning tasks, Terra balancing capability and cost, and Luna optimized for high-volume, fast operations at lower latency and expense. Some tools like Codex still require additional configuration to fully leverage v2 multi-agent functionality.
Source: Read the original article
