Akamai Inference + Agent Sandbox: Give Your AI Agents a Safe Place to Run Code

Akamai launched two products on October 27, 2026, and together they answer the question every team shipping AI agents is quietly asking: where do agents actually run, and what stops the code they write from running wild?

The products are Akamai Inference and Akamai Agent Sandbox. One serves models. The other cages the code agents write. Neither is a new model — and that is exactly why this launch matters for developers.

Disclosure: no affiliate links in this article. Everything below comes from Akamai's October 27 launch announcement. I have not run these products hands-on; this is researched analysis, not a benchmark I performed, and Akamai's claims are Akamai's claims until independent testing checks them.

What Akamai actually launched

Akamai Inference is managed model serving. You get open-weight and custom models through OpenAI-compatible APIs, so existing applications can move without rewrites, according to Akamai. Choose serverless, pay-per-use inference or a dedicated deployment. At launch the catalogue includes 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, plus embedding models for retrieval-augmented generation.

Akamai Agent Sandbox is the more interesting half. Agents run on Akamai Functions, and the code they write runs in its own isolated environment, sequestered from the agent's runtime. It is delivered as an open source component of the Spin framework, and it uses WebAssembly isolation to set clear limits on compute, memory, outbound connections, and system capabilities. In plain terms: the agent can think and act, but the code it generates executes in a box with walls you define.

Both are in beta at launch. Akamai Inference beta includes serverless deployment options. For NVIDIA B300 GPU clusters, Akamai says to contact its sales team — so treat the biggest hardware tier as a sales conversation, not a self-serve click.

Why the sandbox is the real story

Models keep getting cheaper and smarter. That is not the bottleneck anymore. The bottleneck is trust.

The moment an agent writes and runs its own code, you have a program writing programs inside your infrastructure. Without isolation, that code can reach whatever the agent can reach — your data, your network, your credentials, your customers. Akamai's own framing is blunt about this: enterprises moving AI into production need serving infrastructure, access to their data, routing and governance, and, as agents begin to write and execute their own code, a hard boundary around what that code can reach.

This is the same lesson running through the whole agent-infrastructure wave. Replit's business launch this week was built around an agent "habitat" — orchestration, sandboxing, and identity controls. Akamai is now offering its version of that habitat on its own cloud: managed inference on one side, isolated execution on the other, with its security portfolio wrapped around both.

The WebAssembly choice is worth noting. Akamai is not the only isolation model in town — it name-checks microVMs with snapshot and fork capability and policy-enforced containers as alternatives — but WebAssembly isolation is lightweight, fast to spin up, and fits the serverless, edge-shaped workloads Akamai is built for. For short-lived agent code — run a script, transform some data, call an API, return — that is a sensible default. For long-running, stateful, heavily privileged workloads, you will still want to look hard at the microVM and container options Akamai also offers.

What Inference gives developers, practically

Three things stand out from the announcement:

1. No rewrite to switch. OpenAI-compatible APIs mean code written against OpenAI-style endpoints can point at Akamai Inference without a refactor. That lowers the switching cost to near zero for a first experiment — the real cost is in testing behaviour, latency, and output quality on the new models, not in plumbing.

2. Models plus your data, in one place. Akamai Cloud provides managed databases and object storage and connects to private and public clouds, so models can draw on enterprise data. Akamai Inference then handles orchestration, model execution, and authentication. Requests run on Akamai's global network — more than 4,350 edge locations — which Akamai says reduces latency and helps keep performance consistent.

3. A genuinely wide model range. 8 billion to 2.8 trillion parameters is not a token catalogue. The small end is for cheap, fast, high-volume work; the large end — Nemotron 3 Ultra class — is for the heaviest reasoning. Embedding models in the same catalogue mean retrieval pipelines do not need a second provider for the search half of the job.

What Akamai has not published in this announcement: pricing. No per-token rates, no sandbox pricing, no serverless unit costs. For a launch aimed at developers, that is the biggest missing piece. Do not budget on this announcement alone — wait for the pricing page, or talk to sales if you are evaluating at enterprise scale.

The GPU fleet underneath it

Akamai also expanded its dedicated GPU fleet to NVIDIA Blackwell Ultra clusters, and says it is preparing for NVIDIA Vera Rubin in 2027. Each eight-GPU Blackwell Ultra node carries 288 GB of HBM3e memory per GPU — 2.3 TB in total — aimed at test-time scaling, chain-of-thought reasoning, and long-context agentic work.

The practical point for developers: dedicated clusters give you control over model versions and deprecation timelines, and run without rate limits or shared-tenancy queueing, according to Akamai. That matters if you have ever had a provider deprecate the model version your product depends on, or throttle you in the middle of a launch. Managed inference is convenient; dedicated GPUs are control. Akamai is now selling both, on the same platform, which is the whole pitch: build, run, and protect AI in one place instead of assembling serving, execution, and security from three vendors.

Akamai says this AI infrastructure push has already secured multiyear contracts collectively worth more than $14 billion. That is a vendor claim about its cloud business, not a benchmark — but it does tell you this is not a side experiment. Akamai is betting its cloud future on being the place agents run.

Where this fits — and where it doesn't

Use Akamai Inference to evaluate if: you want OpenAI-compatible serving of open-weight models without operating your own GPU stack, you care about edge latency, or you want inference, data, and security on one platform rather than stitched together yourself.

Use Akamai Agent Sandbox to evaluate if: your agents write and run code today — or will next quarter — and you currently have no hard boundary between agent code and the rest of your system. If your agents only answer questions and never execute anything, you do not need this yet. The day they start running code, you do.

Do not switch on this announcement alone if: you need published pricing, independent latency benchmarks, or a production track record. Both products are in beta. Beta means the API, the limits, and possibly the pricing can still move. Pilot it, measure it on your workload, and keep your current provider running in parallel until the numbers are yours, not the vendor's.

What to do this week

1. Audit what your agents can execute today. List every place an agent's output turns into code execution, shell commands, or API calls with write access. Any item on that list without isolation is the argument for a sandbox — Akamai's or anyone else's.
2. Try the OpenAI-compatible endpoint on a non-production workload. Point a test app at Akamai Inference, run the same prompts you run today, and compare outputs, latency, and cost once pricing is published. A compatible API makes this a one-afternoon experiment.
3. Set sandbox limits before you need them. Decide now what agent code may never touch: credentials, production databases, outbound internet, the filesystem outside a scratch directory. Write those limits down, then enforce them in whatever sandbox you choose.
4. Wait for pricing and independent reports before committing. The architecture is sound and the pitch is coherent. The bill is still unknown, and the bill is what decides production use.

The bigger picture

The agent era has a plumbing problem, not a model problem. We have models that can write code. What we have been missing is a boring, reliable place for that code to run — with limits, logging, and a security team standing behind it.

Akamai's launch is not flashy, and that is the point. Inference plus a sandbox plus edge delivery plus security is infrastructure, and infrastructure wins quietly: it becomes the default because it is easier, safer, and cheaper to build on than to assemble yourself. Whether Akamai becomes that default depends on pricing, reliability, and developer experience over the next few months — none of which a launch announcement can prove.

But the direction is unmistakable. The companies winning the agent era will not just be the ones with the smartest models. They will be the ones that give agents a safe place to work. Akamai just entered that race in earnest.

FAQ

What is Akamai Inference? A managed model-serving service launched October 27, 2026, offering open-weight and custom models through OpenAI-compatible APIs, with serverless pay-per-use or dedicated deployments. At launch it includes 12 open-weight models from 8 billion to 2.8 trillion parameters, plus embedding models.

What is Akamai Agent Sandbox? A secure execution environment for AI agent code, delivered as an open source component of the Spin framework. Agents run on Akamai Functions while the code they write runs isolated via WebAssembly, with limits on compute, memory, outbound connections, and system capabilities.

Are Akamai Inference and Agent Sandbox generally available? No — both are in beta at launch. Akamai Inference beta includes serverless deployment options, and NVIDIA B300 GPU cluster availability is via Akamai Sales.

How much do Akamai Inference and Agent Sandbox cost? Akamai's launch announcement does not publish pricing. Wait for the pricing page or contact Akamai Sales before budgeting.

Which models are available at launch? 12 open-weight models including NVIDIA Nemotron 3 Ultra, OpenAI gpt-oss-120b, and Google Gemma 4, plus embedding models for retrieval-augmented generation.

Sources: Akamai's October 27, 2026 launch announcement via GlobeNewswire. All product claims, model counts, and the $14 billion contract figure are Akamai's own and have not been independently verified.