Models

Agents use models through Antigen’s model gateway. Set the model field when assembling an agent, or keep the model configured on a pre-built agent. The gateway provides cyber-capable Claude models alongside GLM and Kimi models. Your Antigen API key provides access through the SDK; you do not need separate provider keys.

Available models

Anthropic’s cyber-capable Claude models are available through Antigen’s model gateway for use against approved targets and vulnerabilities.

Model	Model ID
Claude Opus 5	`claude-opus-5-cyber`
Claude Opus 4.8	`claude-opus-4-8-cyber`
Claude Opus 4.7	`claude-opus-4-7-cyber`
Claude Opus 4.6	`claude-opus-4-6-cyber`
Claude Opus 4.5	`claude-opus-4-5-20251101-cyber`
Claude Sonnet 5	`claude-sonnet-5-cyber`
Claude Sonnet 4.6	`claude-sonnet-4-6-cyber`
Claude Sonnet 4.5	`claude-sonnet-4-5-20250929-cyber`
Claude Haiku 4.5	`claude-haiku-4-5-20251001-cyber`

GLM and Kimi use their standard model IDs without the suffix.

Model	Model ID
GLM 5.2	`zai-org/GLM-5.2`
GLM 5.3	`zai-org/GLM-5.3`
Kimi K3	`moonshotai/Kimi-K3`

List models

Use antigen.models.list() to retrieve the models available to your organization:

import Antigen from "@antigen/sdk";

const antigen = new Antigen(process.env.ANTIGEN_API_KEY);
const models = await antigen.models.list();

for (const model of models) {
  console.log(model.id);
}

Use the returned IDs when configuring agents.

Select a model

Set model when constructing an agent. This example keeps tCell’s configuration and selects Claude Opus 5 Cyber:

import Antigen, { tCell } from "@antigen/sdk";

const antigen = new Antigen(process.env.ANTIGEN_API_KEY);
const original = await antigen.agents.get(tCell);

const agent = antigen.agent({
  ...original,
  model: "claude-opus-5-cyber",
});

Changing the model preserves the other configuration fields, including the agent’s skills, guardrails, and tools. To change the model saved on a registered agent, use antigen.agents.update():

await antigen.agents.update("our-security-agent", {
  model: "claude-opus-5-cyber",
});

See Agents for constructing and updating agent configurations.

Model access during runs

Cyber-capable models are available to agents running in sandboxes, where target restrictions control which systems they can test. For offensive testing, submit your targets for approval before starting a run. See Targets for submitting targets and checking their status.

Getting started

Asset Map

Runs

Vulnerability Lifecycle

Team Workflows

SDK

CLI

Integrations

Available models

List models

Select a model

Model access during runs

​Available models

​List models

​Select a model

​Model access during runs

Available models

List models

Select a model

Model access during runs