Dedicated Agent OS servers, unmetered local inference, and one OpenAI-compatible key for the cloud models your agents need.
Choose a complete Agent OS command center, dedicated Windows compute, a VPS with model access, or a reserved GPU lane.
Each standardized card follows the same rhythm: compute, model allowance, and only the new capabilities unlocked at that tier.
Agent OS, a practical VPS, and the first cloud model set.
Included with STARTER:
More local speed and a broader model pool for growing fleets.
Everything in STARTER, and also:
The complete dedicated Agent OS starting point.
Everything in GROWTH, and also:
More memory for a larger command center.
Everything in EXTREME, and also:
Production capacity for parallel agent work.
Everything in MONSTER, and also:
High-memory parallelism for fast fleets.
Everything in BEAST, and also:
Maximum context room for aggressive workflows.
Everything in LIGHTNING, and also:
The full catalog and maximum local scale.
Everything in DRAGON, and also:
Flagship credit is separate from the pooled base-model allowance.
Latest-generation AMD EPYC resources with administrator access and a 100 Gbit network.
OpenAI-compatible API keys with a monthly token allowance and guaranteed throughput. Pay monthly, or pay yearly and save.
Monthly and annual options are available for eligible plans.
Reserve single-GPU capacity today or talk with the team about a custom multi-GPU deployment.
Filter by the first Agent OS tier that unlocks access. Higher tiers inherit everything below.
26 models available
Grok 4.6 is the flagship lane and uses explicit credit rather than the pooled base-model allowance.
Dedicated compute, fast storage, high-throughput networking, and a unified model-access layer arrive as one environment.
Choose Windows or Ubuntu with the command-center stack ready for your team.
Keep high-volume loops on local inference without per-token charges.
Protect agent state with environment snapshots designed for recovery.
Your operating environment and proprietary workflows stay isolated.
High-capacity networking supports concurrent research and data-heavy workflows.
Keep the same client interface when your plan or model choice changes.
Routine work stays on the unmetered local lane. Specialist calls route through the API catalog included with your tier.
Pick Agent OS, Windows, API + VPS, or dedicated GPU capacity.
Your server resources and access arrive scoped to the selected tier.
Keep recurring agent loops on your dedicated local inference lane.
Use the compatible key when a workload needs a cloud model.
From a legacy machine to a multi-GPU cluster, blade system, or full cabinet, talk directly with the team about capacity, networking, and deployment.