Anthropic launched Opus 5.5, claiming roughly 40 percent savings versus Opus 5 for typical workloads at default settings due to lower token costs and reduced token usage. The model excels in high-risk areas like cybersecurity and biology, retaining Fable 5.1 protections that may route flagged requests to older models. OpenAI released GPT-6 Sol and Luna as cost-focused upgrades trained with methods similar to GPT-6 Astra.

Sol targets efficient daily-driver tasks while Luna offers a fast, cheap option. Both cost half as much as predecessors and show modest capability gains of a few percentage points on certain benchmarks. OpenAI's naming convention positions Astra as the premium model, Sol as the efficient alternative, Terra as balanced general-use, and Luna as the budget tier.

Rough comparisons map Astra to Fable, Sol to Opus, Terra to Sonnet, and Luna to Haiku, though the mapping is imperfect for higher-end models.