How to get the most from your Claude token usage

By Todd Heater · Published 9 July 2026 · A practical guide for busy UK bid teams

Most people never need the expensive Claude plan. With a little rhythm, a standard Claude Pro subscription plus a small amount of extra usage will carry a working bid team through weeks of drafting. The trick is to work with the way Claude measures usage, not against it. Here is how we do it.

How Claude's usage limits actually work

Claude measures your usage in two layers, and each resets on its own clock.

The first is a rolling five-hour session. Your first message of the day opens a five-hour window, and everything you do until it closes counts against that window's allowance. Five hours after that first prompt, it resets. Your chat and your Cowork work draw from the same pool, so the session covers all of it.

The second is a weekly ceiling. On top of the five-hour rhythm there is a weekly cap. You will rarely meet it in ordinary use, but it is worth knowing about when you are drafting a big bid across several days.

You can see exactly where you stand at any time under Settings > Usage, which shows a bar for your current five-hour session and a bar for the week. Anthropic raised the five-hour allowances in 2026, so there is more room than there used to be. Once you can see the clock, you can plan around it. Here are four ways to do that.

1. Time your day around the reset

The single biggest win is to start the five-hour clock before you actually need it.

If your day starts at nine, make your very first action a small one: kick off a short conversation, ask Bidwin Watch for the morning's opportunities, or clean-copy a draft from the night before. That first prompt starts the countdown. Five hours later your session resets, right about when you are ready for the heavy afternoon drafting, so you go into the big work with a full allowance rather than a half-spent one.

Use the same logic at the end of the day. When you sign off, your session has usually reset again, so that is the moment to launch a large job you have queued up: a full first-draft batch, a long clean-copy pass, a whole tender vet. Set it running and let it work overnight. Just make sure your computer will not go to sleep while it does, or the job stalls when the screen locks.

One honest caveat. The five-hour trick paces your day, but the weekly ceiling paces your week. If you have a marathon bid on, spread the biggest batches across a few days rather than burning everything on Monday.

2. Right-size every job

You would not use a sledgehammer to hang a picture. Match the effort to the task and you will get far more done inside the same allowance.

Pick the right model for the job. Draft routine sections with a faster, lighter model, and step up to a stronger one only for the complex, high-stakes questions. Bidwin recommends the right setting as you go, so you do not have to track model releases yourself.

Use Quick Vet when you do not need the full picture. Bidwin's tender vet has a lighter mode that leads with the question inventory, evidence map and target lengths, and skips the fuller analysis. It is quicker and cheaper when that is all you need.

Save extended thinking for the hard parts. Thinking is powerful but it costs more, so leave it off for routine editing and turn it on for the genuinely tricky sections.

Batch when you have room, and go one section at a time when you do not. On a tight allowance, draft a single section. When you have a fresh window, and overnight is ideal, use Max batch to draft as many sections as safely fit in one sitting.

3. Keep your context lean

Everything sitting in a conversation gets re-read on every turn, so a bloated context quietly costs you tokens all day. Keep it tight.

Attach only the files a task needs. Point Bidwin at the relevant evidence for this question, not your entire library. A smaller context is cheaper and faster, and it is better for confidentiality too.

Start a fresh task for a fresh job. A single endless conversation carries all of its history forward and charges you for it again and again. When you move to a new tender or a new question, start a new task. It is cleaner, and it saves tokens.

Have your files ready before you run a skill. Make sure your House-Style file is in place so setup runs fast, and use Claude to prepare an instruction file for the new tender so Quick Vet has what it needs. A skill that has to stop and ask, or hunt for a missing file, is a skill that costs you a re-run.

Keep your files local and downloaded. Cloud files that have not fully downloaded can stall a job midway. Bidwin checks for these before it starts, but keeping your working folder local avoids the problem altogether. And reuse rather than repeat: your drafting plan and house style are built once and used across every section, so there is no need to vet the same tender twice.

4. Never let a job get blocked

The worst way to lose time is a job that breaks halfway because you ran out of allowance, right when you were ready to start your own human review.

Keep extra usage standing by. Claude lets you turn on extra usage, so that when you reach your plan's limit you can keep going and simply pay for what you use, rather than stopping dead. We always keep it enabled. When a complex job would otherwise break, a pound or two to see it finish is money well spent.

But wait when waiting is free. Check Settings > Usage before a big run. If you are near your limit and the reset is only minutes away, put the kettle on and let it reset rather than paying. If it is hours away and you need the work now, that is exactly when extra usage earns its place.

Set your model before you start. Cowork locks the model once you send the first message of a task, so choose it up front. Starting a big job on the wrong model wastes the whole run.

An example working day

At nine, start the clock with something small, such as a Bidwin Watch request or a clean-copy of last night's draft. Late morning, prepare the day's tender: check the house style is in place, build an instruction file, and run a Quick Vet. By early afternoon your session has reset, so the heavy drafting gets a full allowance. Through the day, glance at Settings > Usage now and then; if you are near the limit with a reset minutes away, wait, and if you need it now, use extra usage. At the end of the day your session has usually reset again, so launch the big overnight batch you queued up, keep the machine awake, and let it run.

Do you need Claude Max, or is Pro enough?

You do not have to jump to the big plan to work this way. Start on standard Claude Pro, around £16 to £18 a month, with extra usage switched on for the busy days. Run like that for a while. If you find you are regularly leaning on extra usage and pushing your weekly ceiling, then, and only then, the higher Max tier starts to pay for itself. Buy the capacity when your own usage proves you need it, not before.

Frequently asked questions

Claude's limits and prices change from time to time. Check Settings > Usage for your live figures, and Anthropic's help centre for the current plans. Last reviewed 9 July 2026.

← Back to all resources