Snowflake adds GPT-6 Sol and Luna to Cortex Inference in public preview

Snowflake has made OpenAI’s GPT-6 Sol and GPT-6 Luna available in public preview through Cortex Inference, giving customers a Snowflake account endpoint for calling the models.

Cortex Inference follows the OpenAI Chat Completions specification and works with OpenAI’s Python and JavaScript SDKs. Snowflake says inference runs within its security and governance perimeter. Its documentation marks openai-gpt-6-sol and openai-gpt-6-luna as public-preview models and says preview models are not suitable for production workloads. Access can vary by region and account configuration.

Snowflake’s announcement describes support in Cortex AI Functions as coming soon. Snowflake’s current Cortex AI Functions availability table separately lists both model IDs under AI_COMPLETE with the private-preview symbol, which it defines as access limited to explicitly enabled accounts. That listing does not establish public availability or a delivery date. For a comparison of the same Snowflake channels in another model release, see DataPhoenix’s Grok 4.6 Cortex AI rollout.

OpenAI released GPT-6 Sol and GPT-6 Luna on September 22 for its Responses and Chat Completions APIs, with text and image input and text output. Snowflake characterizes Sol as the stronger option for complex professional and technical work and Luna as designed for responsive execution at scale.

Support in Snowflake CoCo, Snowflake CoWork, and Cortex Agents remains described as coming soon. Snowflake has not provided delivery dates for those integrations, and its announcement says forward-looking product statements are not commitments to deliver.