OpenAI has announced two new models, GPT-6 Sol and Luna. According to the company, both models offer lower cost and produce fewer mistakes. OpenAI describes the pair as being derived from the same lineage as its Astra model.
Why it matters
Lower operating cost and reduced error rates are two of the most commonly cited priorities for organizations deploying large language models, so any change on these dimensions could affect how the models are used.
Who should care
Developers and organizations that build on OpenAI’s models may want to evaluate GPT-6 Sol and Luna against their current setups. Note that the specifics reported here are limited to the company’s own claims.