Pricing was the part I really didn’t want to think about. I’m a developer — I’d much rather build the thing than sit there agonising over the number. But pricing compounds more than almost anything else, so I forced myself to do the work.
The market’s already settled around $20/month — ChatGPT Plus, Claude Pro, all of them. That’s a real anchor.
T3 Chat — an AI chat app running at $8/month — showed the price point can work, but only with care. Its founder Theo Browne has been open about what happened — he picked the price before adding Claude, and ended up with individual users costing him over $200 on $8 plans. The fix wasn’t to raise the price. It was to add proper limits.
I’d been reading Priceless by William Poundstone. If you’re pricing anything, read it. People don’t evaluate prices on a perfectly smooth scale. They bucket them. £9.99 feels like £10. £7.99 feels like it belongs in a lower bracket. That last digit does more work than you’d expect.
The maths worked at several price points. The question was which one felt right.
£10 was the top end for me. Once you cross into double digits, it feels meaningfully more expensive, even if the difference is tiny. £5 felt too cheap. Not because the maths didn’t work, but because the signal felt wrong. If I want people to take Layzer seriously, charging less than a pint makes it sound like I don’t believe in it.
That left £6 to £9. £7.99 felt a bit too clever, a bit too obviously like a marketing trick. £8.00 felt cleaner. Deliberate. Above the bargain zone, below the point where people start flinching.
What makes £8 survivable is the credit system:
| Tier | Credits | Refills | Standard models | Premium models |
|---|---|---|---|---|
| Free | 500 | every day | ~10–15 messages | not available |
| Pro | 200,000 | every month | ~4,000 messages | ~150 messages |
A quick note on how I got there. I worked out the averages assuming a typical message is around 2,000 input tokens (your prompt plus a bit of chat history) and 500 tokens out. That’s enough for a normal back-and-forth — long conversations cost more, short follow-ups cost much less. Costs come straight from each provider’s published per-token rates.
| Model | Tier | Credits per message |
|---|---|---|
| GPT-5 Nano | Standard | ~30 |
| Llama 4 Scout | Standard | ~31 |
| DeepSeek V4 Flash | Standard | ~42 |
| Claude Sonnet 4.6 | Premium | ~1,350 |
Free is a taster — most people will hit the daily cap, and that’s by design. Pro is where Layzer actually lives. The users who would have cost me $200 a month on a flat-rate plan can’t get there, because the credits run out first.
Two tiers, Free and Pro. I’ll add more when I have a reason to.
Will it hold forever? Probably not. I expect power users will push it, costs will creep, and I’ll probably need a higher tier within six months. But I keep coming back to the same idea: you don’t raise the base price unless you really have to. You add a tier above it.
So £8 it is.