Per-character pricing run against flat streams.
15,000 min
where our own flat stream passes our own tokens
A lab report on a pricing model, not on a vendor. Published meters run $30 to $100 per million characters, a spoken minute is about 1,000, and here the same speech is $10 a million prepaid or $150 a month on a flat stream.
The crossing, drawn
One month of one stream, both models on the same axes. The meter is a wedge, because published rates span $30 to $100 per million characters; the line is a horizontal at $150.
fig.
One stream, one month, meter wedge vs flat stream
Fig
Meter cost = minutes × 1,000 chars × rate ÷ 1M, at the published $30, $100 range (July 2026). The flat rate crosses the cheapest meter near 5,000 spoken minutes and the priciest near 1,500. Our own tokens sit below the whole wedge at $10 a million, and the flat stream passes them at 15,000 minutes.
The arithmetic
Every cell is one multiplication. Check any of them. The token column is minutes × 1,000 characters × $10 ÷ 1M, and it is under the meter in every row, because $10 a million is below the bottom of the published range.
The only crossing that decides anything is between our own two doors. Take tokens up to 15,000 spoken minutes a month and the flat stream past it. The last row is the flat stream winning, and the first three are the pack.
table 01
Four workloads, one month, three models
| Workload | Minutes / mo | On the meter | Gandr tokens | Gandr flat |
|---|---|---|---|---|
| Reminders only | 2,000 | $60, $200 | $20 | $150 |
| E-learning library | 6,000 | $180, $600 | $60 | $150 |
| Audiobook catalogue | 12,000 | $360, $1,200 | $120 | $150 |
| Dubbing catalogue | 30,000 | $900, $3,000 | $300 | $150 |
Run it on your own traffic
Take one month of spoken minutes out of your own logs and multiply by 1,000 characters. That count times $10 a million is your bill on tokens here; your stream count times $150 is your bill on flat. Whichever of the two is smaller is the door to take, and both are on the same engine.
A key and one stream, and the table above becomes something you rerun against your own traffic.
Get a key and hold us to these numbers