OpenAI Group PBC today announced plans to make its latest large language models accessible in Google Cloud.
The LLMs will be available through the Alphabet Inc.’s unit Model Garden service. It provides access to more than 200 neural networks from Google, its partners and the open-source ecosystem. The lineup includes not only LLMs but also smaller, more specialized models optimized for tasks such as translation.
The first LLM OpenAI is bringing to Google Cloud is its flagship GPT-6 Astra model. The algorithm is responsible for several of the AI-produced math discoveries that the company has published in recent weeks. It also achieved a perfect score on ExploitBench, a benchmark that measures LLMs’ ability to find and exploit cybersecurity flaws.
Astra includes safety optimizations designed to make it suitable for enterprise use. According to OpenAI, the model causes “fewer misaligned outcomes” than all the other frontier models it has evaluated. It’s configured not to bypass user-defined restrictions on the tasks that it can perform.
Google Cloud customers will also receive access to GPT-6.1 Sol and GPT-6 Luna, two scaled-down versions of Astra that OpenAI launched in late September. They trade off some output quality for a significant reduction in token pricing. Sol and Luna cost five and 100 times less than Astra, respectively.
“We are focused on giving customers a broad selection of models, including those from OpenAI, so they can choose the right model for any given job, helping to deliver optimal quality and lower costs,” said Google Cloud Chief Executive Thomas Kurian.
Cloud providers offer discounts to enterprises that sign long-term spending commitments. OpenAI says that Google Cloud customers can use those commitments to pay for access to its models. That removes the need for users to set up a new AI budget, which eases procurement.
Astra, Sol and Luna will become generally available in Google’s Model Garden within a few weeks. The companies plan to follow up the launch by bringing the models to the search giant’s Confidential Compute portfolio of infrastructure services.
Server-grade central processing units such as those Google uses include a security feature called TEE. The technology can encrypt a segment of the host server’s memory in a manner that only makes it accessible to a specific workload. Google’s Confidential Compute services use the technology to secure customer workloads. Additionally, they can ease use cases such as sharing a dataset among multiple organizations for AI training purposes.
OpenAI also plans to make its models available via Google’s Antigravity and Gemini Agent products. Antigravity is a vibe coding tool that uses agent ensembles to speed up software projects. Gemini Agent, in turn, is designed to automate business tasks that span multiple cloud applications.