A bigger commit earns a bigger discount, and that headline is true. What the headline leaves out is the math on the part of the commit you never use. Do that math and the effective rate often looks nothing like the rate on the quote. Here is the calculation the seller would rather you not run.
The pitch for a larger Claude commitment is simple and seductive. Commit more, and the per token rate drops. The quote shows a clean discount off list, the savings look real, and the bigger the number you commit, the better the headline rate gets. All of that is true as far as it goes. What the quote does not show, and what the account team has no reason to volunteer, is the math on the portion of the commit you do not consume. On most Anthropic agreements, unused commitment is forfeited. It does not roll into the next period and it does not come back as credit. So the rate that matters is not the rate printed on the quote, it is the effective rate you actually pay once the unused portion is counted as money spent for nothing. That number is the one the seller hopes you skip, and it is the only one worth deciding on.
Here is the calculation in plain terms. Suppose you commit to a level of spend and earn a discount that makes the headline rate look attractive. If you consume the full commitment, the headline rate is the real rate and the deal is exactly as good as it looks. But if you consume only part of it, you still pay the whole commitment, so your effective rate is the committed dollars divided by the tokens you actually used. The fewer tokens you use against a fixed commitment, the higher your effective rate climbs, and it can climb right past the list price you were trying to beat. A discount of twenty five percent on paper can become a premium over list in practice if you used only seventy percent of what you committed. The discount was real on the tokens you consumed and a total loss on the tokens you did not, and the blended result is what you actually paid.
When you put those three side by side, the decision gets clear. If the effective rate at the low end of your range still beats list, the commit is genuinely good even in a bad year. If the effective rate at the low end is worse than list, you are betting on optimistic consumption to make the deal work, and that is exactly the bet the structure is designed to win for the seller. The whole point of running the math is to stop deciding on the headline and start deciding on the effective rate in the case you are actually worried about.
This is not an accident of contract drafting, it is the economics of committed spend working as intended. A commitment transfers consumption risk from the seller to you. In exchange for a lower rate, you guarantee a revenue floor, and if your usage falls below that floor you make up the difference out of your own pocket through forfeited commitment. That is a reasonable trade when you are highly confident of your usage, and a poor one when you are not. The larger the commit relative to your confident usage, the more risk you have taken on and the worse the deal becomes in any year that disappoints. The seller is happy to grow the commit because each additional committed dollar is either consumed at a discount or forfeited at full value, and both outcomes are good for them. Your job is to commit only up to the level where you are confident the consumption is real, and to handle everything above that as protected overage rather than a pre paid bet.
The remedy is not to avoid committing, because a right sized commit at a real discount is a good deal. The remedy is to size the commit to the consumption you are confident about and to negotiate the structure that protects the rest. Commit near the lower end of your honest range, where the effective rate and the headline rate are close because you will use almost all of it. Then secure the right to consume above the commit at the committed rate, so the upside you were not sure about gets the discount without you having pre paid for it. Add a phased ramp if your usage is growing, so the commit steps up with proven consumption rather than starting at a number that assumes a year of perfect adoption. Each of these turns the asymmetry back in your favor by making sure the dollars you commit are dollars you will actually spend.
The other lever is to make the consumption itself smaller and more predictable before you commit to anything, because a smaller honest forecast means a smaller commit and a smaller gap to forfeit. Routing predictable work to Sonnet and Haiku instead of running everything on Opus pulls aggregate spend down sharply. Prompt caching on repeated context can cut the cost of that context by up to ninety percent, and batch processing on asynchronous work runs at roughly half rate. Doing this work first means the number you take into the commit conversation is already lean, so the discount you negotiate sits on top of real efficiency rather than papering over waste. A commit sized to an unoptimized workload locks in your inefficiency at a discount, which is a worse outcome than it sounds.
One specific version of this math deserves naming, because it catches careful buyers. The seller offers a deeper discount if you step the commit up to the next band, and the per token rate at the higher band genuinely is better. The trap is comparing the two headline rates while ignoring the consumption gap you would be creating. If reaching the higher band means committing well above your confident usage, the extra committed dollars are at risk of forfeiture, and a deeper rate on dollars you forfeit is no saving at all. Run the effective rate at both bands against your honest low case. Often the smaller commit at the shallower discount wins, because you actually consume all of it, while the larger commit at the deeper discount leaves a forfeited tail that drags the effective rate above where the smaller commit landed. The better headline rate lured you into a worse real outcome. The only way to see it is to do the effective rate math at both bands rather than comparing the printed discounts.
The entire problem rests on one contract term that the quote rarely highlights: what happens to commitment you do not use. On most Anthropic agreements the answer is that it is forfeited, but the treatment is negotiable like anything else, and a buyer who raises it early can sometimes improve it. Ask directly whether unused commitment can roll forward into a later period, whether a portion can carry over, or whether a shortfall can be made up by consumption in the following term rather than written off. Even partial relief here changes the effective rate math in your favor, because it shrinks the cost of the gap between commit and actual usage. The seller has no reason to volunteer this, so the buyer has to put it on the table. Pairing a conservative commit, overage at the committed rate, and some form of carryover on the unused portion is how you turn the structure from a one sided bet into a fair arrangement.
We sit between you and Anthropic and we run the effective rate math before you sign, not after the year disappoints. That means sizing the commit to your confident consumption, computing the real rate at the low end of your range, and negotiating overage at the committed rate plus a phased ramp so the structure protects you instead of the seller. Alongside that we apply the optimization levers, routing across Opus, Sonnet, and Haiku, caching at up to ninety percent on repeated context, and batch at roughly half rate, so the commit is smaller and the gap that could be forfeited is smaller still. The result is a deal where the rate you pay and the rate on the quote are close, because the commit was sized to be consumed.
If you have a commit quote in hand and you want the effective rate math run before you sign, that is the fastest, highest value conversation we have. Our pricing is simple: a Fixed Fee from $18,000, or Gainshare which is a share of verified savings with zero retainer and no risk to you. Get a quote and we will run the numbers the seller hoped you would skip.
Get a quote for a bounded engagement. Fixed fee or gainshare, no risk to you.
Get a QuoteWeekly intelligence on Anthropic pricing moves and the buyer side counters that work.