Four names for four different jobs. Picking the wrong tier doesn't just cost more than necessary — for long-context work, it can directly return the wrong answer.
Sol is the flagship for hard coding, agents, and research — $5/$30 per million tokens. Terra is the balanced option, with quality close to GPT-5.5 at less than half the price — $2/$12 since the July 30, 2026 price cut. Luna is the cheapest, optimized for latency — $0.20/$1.20 after an 80% price cut.
Sol scores 61 points at maximum effort on the independent Artificial Analysis index, and leads the Coding Agent Index with 80 points — for complex coding and agentic tasks, the gap versus Terra and Luna isn't marginal, it's substantial.
Sol and Terra land nearly tied on long-context tasks (91.5% and 89.6% respectively), while Luna drops to 41.3% — a significant performance cliff for document analysis and multi-document synthesis. If your workload involves long-context retrieval (document analysis, reasoning over large codebases, multi-document synthesis), Luna is the wrong tool, no matter how much cheaper it is per token.
Terra is the sensible default for most work. Sol for the hardest agentic tasks. Luna for high-volume pipelines — but only once you've confirmed the task doesn't depend on long context, given the performance cliff documented above.
GPT-5.6-Cyber is a model trained specifically for cybersecurity, built on top of Sol, available only through OpenAI's Daybreak Red trusted-access program — no public API and no published price. We already covered the Daybreak program in detail: this tier isn't a standard catalog option, it's restricted, vetted access for legitimate offensive and defensive security use cases.
Carlos Montiel is an enterprise AI solutions architect. He implements LLMs, Agents, RAG, and orchestrators for companies across Guatemala and Latin America. Reach out for a consultation.
Contact Carlos Montiel