Enterprise Pricing and Negotiated Tiers

Updated November 2, 2026 — flagship model. Grok 4.5 is no longer the most capable model: Grok 4.6 (August 12, 2026) and then Grok 4.7 (grok-4.7, September 21, 2026) replaced it, with 500,000 context tokens, reasoning effort from low to xhigh, and the same tiered pricing ($2 / $6 under 200,000 prompt tokens, $4 / $12 above). grok-4.5 remains available; for a new project, start with grok-4.7 (course Grok in 2026, lesson 9).

Two Pricing Logics

xAI Enterprise pricing doesn’t extend the pay-as-you-go model; it changes its nature. In standard access, public rates apply per token with no commitment: your bill follows your activity faithfully. These rates are published in the official documentation, vary by model, and always distinguish input tokens from output tokens; an automatic caching mechanism reduces the amount when the same prompt prefixes are reused from one request to another.

The Enterprise tier substitutes this mechanism with individual negotiation, where conditions depend on your volumes and needs. Discussions most often focus on volume discounts proportional to your monthly commitment, locked rates guaranteeing a price for the contract duration regardless of public changes, hybrid plans combining provisioned throughput and pay-as-you-go, or included credits—a volume of tokens offered for startup or testing. Locked rates deserve particular attention: for a finance director budgeting over three years, price stability sometimes weighs heavier than the absolute level.

Structuring the Commitment

Volume commitment is the most direct formula: you commit to a monthly minimum of tokens and xAI applies a discount on all your consumption in return. This usage must be predictable. A company whose application traffic varies little from month to month benefits; a highly seasonal activity gets locked in and pays for tokens it won’t consume in February.

Provisioned throughput commitment addresses another concern. You reserve capacity rather than volume, billed per unit per day. What you buy isn’t a price per token, it’s consistency: stable performance regardless of the platform’s overall load, which becomes critical as soon as a critical application depends on response times.

That’s why the hybrid model wins in most Enterprise deployments. A company’s usages don’t all have the same criticality, and there’s no reason to treat them the same way. Provisioned throughput covers applications that can’t slow down—customer support, real-time analysis—while pay-as-you-go absorbs secondary usages like batch analysis or experimentation. This distribution isn’t just contractual: the API headers x-pt-disable and x-pt-id let you dynamically route each request to dedicated capacity or shared capacity, letting you arbitrate request by request rather than application by application.

Optimizing the Bill

The most effective lever, by far, is to use the right model for each task. The Grok range offers widely spaced price/performance compromises. The grok-4.5 model is the most powerful and justified for complex tasks requiring advanced reasoning. The grok-4.3 model presents an excellent quality/price ratio for common tasks and remains the most economical of the generalist models: it suffices for classification, summarization, and simple processing. The grok-build-0.1 model, specialized for code, displays the lowest rate in the range and suits development workflows. Classifying ten thousand tickets per day with the most powerful model means paying for reasoning capacity the task has strictly no use for.

Caching, meanwhile, demands only writing discipline. The xAI API automatically reduces the cost of input tokens when the prompt prefix is identical from one request to another. So place system instructions and context at the start of the prompt, vary only the end—typically the user’s question—and group similar requests to maximize the cache hit rate. An internal assistant whose system prompt runs two thousand tokens and the question fifty thus sees most of its input cost melt, provided the order of blocks never changes.

Then there are prepaid credits, purchasable via the Management API or the xAI console, which sometimes come with advantageous conditions. A reserve applies here: these credits are non-refundable, except by legal obligation. So prepay only the portion of consumption you’re certain of, and leave the rest on classic billing.

Tracking What You Spend

The Enterprise offer provides the instruments necessary for these arbitrages to rest on figures rather than impressions. Usage analysis details consumption by model, by day, and by API key—it’s what will reveal that a forgotten test environment is absorbing a quarter of your budget. Cost preview gives a real-time estimate of the current bill, spending limits set configurable ceilings that avoid overages, and detailed invoices keep a history broken down by usage. Configure these limits as soon as the account activates: a limit set after the first billing incident always comes too late.

Key Takeaways

• Enterprise rates are negotiated individually and can include volume discounts
• The hybrid model (provisioned throughput + pay-as-you-go) is the most common in production
• Choosing the right model per use case is the main lever for cost optimization
• Prepaid credits are non-refundable
• The Management API offers detailed tracking of consumption and billing