{"slug": "gpt-5-6-has-72-possible-configurations-what-s-a-good-default", "title": "GPT 5.6 Has 72 Possible Configurations. What's A Good Default?", "summary": "OpenAI's GPT 5.6 release offers 72 possible configurations across model size, reasoning effort, task mode, and speed settings, leaving users to determine optimal defaults. The combinations span three model variants (Sol, Terra, Luna), six effort levels, and toggles for Work/Codex and Standard/Fast modes.", "body_md": "# GPT 5.6 Has 72 Possible Configurations. What's A Good Default?\n\nYes, there are probably too many options to choose from in the GPT 5.6 release.\n\nIn the context of reasoning models, though, I find it interesting how these choices map onto training-time and [inference-time scaling](/glossary/#inference-time-scaling). If we loosely map the options onto the classic o1 plots, Sol, Terra, and Luna stand in for three model sizes and training budgets along the training-compute axis. The effort settings then sit on the inference-time-compute axis.\n\nThe full list has three model choices and six reasoning-effort levels (Light, Medium, High, Extra High, Max, and Ultra). Once Work versus Codex and Standard versus Fast are included, the full configuration matrix becomes\n\n`Work/Codex × Sol/Terra/Luna × Light/Medium/High/Extra High/Max/Ultra × Standard/Fast`\n\nThat gives us `2 × 3 × 6 × 2 = 72`\n\npossible configurations.\n\nSo, what is a good default now? Luna with High effort? Sol with Light effort? Terra with Medium effort?\n\nSure, a performance-versus-cost chart can help identify the good bang-for-the-buck combinations. For instance, Luna with Extra High effort may be better and cheaper than Sol with Medium effort.\n\nBut yeah, 72 possible configurations leave us with a lot of choices 🤯.\n\n## Read Next\n\n[Inkling: A New Open-Weight 975B MoE with a Few Surprises Short note on Thinking Machines Lab's 975B Inkling open-weight model, its benchmark profile, sparse MoE design, short convolutions, embedding RMSNorm, and](/blog/2026/inkling-architecture-benchmark-notes.html)\n\n[200,000 Subscribers Short note celebrating Ahead of AI reaching 200,000 subscribers.](/blog/2026/ahead-of-ai-reached-200000-subscribers.html)\n\n[Build a Reasoning Model From Scratch Is Out Short note announcing the release of Build a Reasoning Model From Scratch and linking the publisher and Amazon pages.](/blog/2026/build-a-reasoning-model-from-scratch-is-out.html)", "url": "https://wpnews.pro/news/gpt-5-6-has-72-possible-configurations-what-s-a-good-default", "canonical_source": "https://sebastianraschka.com/blog/2026/gpt-5-6-configurations.html", "published_at": "2026-07-09 22:33:25+00:00", "updated_at": "2026-07-16 15:01:31.231973+00:00", "lang": "en", "topics": ["large-language-models", "artificial-intelligence", "ai-products", "ai-infrastructure"], "entities": ["OpenAI", "GPT 5.6", "Sol", "Terra", "Luna"], "alternates": {"html": "https://wpnews.pro/news/gpt-5-6-has-72-possible-configurations-what-s-a-good-default", "markdown": "https://wpnews.pro/news/gpt-5-6-has-72-possible-configurations-what-s-a-good-default.md", "text": "https://wpnews.pro/news/gpt-5-6-has-72-possible-configurations-what-s-a-good-default.txt", "jsonld": "https://wpnews.pro/news/gpt-5-6-has-72-possible-configurations-what-s-a-good-default.jsonld"}}