How our pricing works
Meridian charges per token — no subscriptions, no seat minimums, no hidden fees. You pay for exactly what you use, plus a flat 20% margin that keeps the lights on and the GPUs spinning.
The formula
Raw cost is what the upstream model provider charges us per token. We multiply by 1.20 — a 20% margin — and pass the result straight to you. No rounding games, no blended rates.
Per-token raw cost
Every model has a published per-token price. We surface it transparently so you can audit every charge.
| Model | Input / 1M tokens | Output / 1M tokens | You pay (×1.20) |
|---|---|---|---|
| GPT-4o | $2.50 | $10.00 | $3.00 / $12.00 |
| Claude 3.5 Sonnet | $3.00 | $15.00 | $3.60 / $18.00 |
| Gemini 1.5 Pro | $1.25 | $5.00 | $1.50 / $6.00 |
| Llama 3.1 405B | $2.00 | $6.00 | $2.40 / $7.20 |
Prices shown are illustrative. Live rates are always visible in your dashboard and may change as upstream providers adjust their pricing. We never change your rate mid-request.
No subscription. No commitment.
You add credits to your account and they sit there until you use them. Credits never expire. There is no monthly minimum, no seat license, and no penalty for pausing. If you stop sending requests, your balance stays exactly where it is.
- ▸Top up with any amount — $5, $500, or $5,000.
- ▸Credits are drawn down per-request at the 1.20× rate.
- ▸Low-balance alerts via email so you never hit a hard stop.
Want to see real numbers for a typical chat session or a batch summarization job? We have worked examples with exact token counts and final costs.