Akamai launches managed AI inference and agent sandbox
Akamai Inference offers 12 open-weight models via an OpenAI-compatible API, while Agent Sandbox isolates AI-generated code.
Akamai has launched Akamai Inference, a managed AI inference service on its cloud, alongside Akamai Agent Sandbox, an isolated execution environment for AI agents. The company announced both on October 28, 2026, according to digitaltoday.co.kr.
Akamai Inference provides managed access to open-weight and custom models through an OpenAI-compatible API, so developers can use it without rewriting existing applications. Customers can pick between serverless pay-as-you-go and dedicated deployment.
At launch, the service offers 12 open-weight models ranging from 8 billion to 2.8 trillion parameters. The lineup includes NVIDIA Nemotron 3 Ultra, OpenAI gpt-oss-120b and Google Gemma 4.
Agent Sandbox runs code written by AI agents in a separate space isolated from the agent itself. Isolation is WebAssembly-based and restricts compute, memory, external connections and system permissions.
The sandbox is built on the Fermyon Spin framework, which Akamai acquired along with Fermyon. Akamai said it plans to add controls over what agents can access and what can access agents.
Quick answers
What models does Akamai Inference offer?
At launch, 12 open-weight models ranging from 8 billion to 2.8 trillion parameters, including NVIDIA Nemotron 3 Ultra, OpenAI gpt-oss-120b and Google Gemma 4.
How is Akamai Agent Sandbox isolated?
It uses WebAssembly-based isolation that restricts compute, memory, external connections and system permissions, running AI-written code separately from the agent.