Pricing.
You pay the per-token prices on each model, plus a service fee of 3.15% + 30c on each top-up. No required subscription, no seat fee, no charge for keys or logs.
| Tokens | Per million tokens, as listed on each model page |
|---|---|
| Service fee | 3.15% + 30c on each top-up, shown before you pay |
| Sub-keys | Unlimited, free |
| Request logs | 90 days, free |
| Minimum top-up | $5 |
| Subscription | None |
| Seat fee | None |
How pricing works
Every model has a price per million tokens for input, cached input and output. The figures on this page and the models page are the prices you are billed, so the price you read there is the price on the bill. You buy credit in advance, and each top-up carries the service fee of 3.15% + 30c, shown before you pay and added on top of the credit.
Billing is in whole nano-dollars, a billionth of a dollar, so short calls are costed exactly rather than rounded up into a bigger number.
Spend caps are part of the price
Every key can carry a lifetime cap, a monthly cap, or both. MagmaRouter reserves the worst-case cost of a request before it leaves and clamps the response length to what the cap covers, so a key cannot bill past the number you set on it. A request that will not fit under the cap is refused instead of sent.
That matters most for agents: something looping overnight is bounded by its key, not by your whole balance. The mechanics are on the spend caps page.
Live per-model prices
Prices differ by model, and every price below is per million tokens. Input is a fresh prompt token, cached input is a prompt token already in the cache, and output is what the model writes. The models page lists every model reachable through one key with its context length and its price, and the catalogue is updated regularly. Point your client at https://api.magmarouter.com/v1 and the model id decides the cost.
| Model | Input | Cached input | Output | Context |
|---|---|---|---|---|
| GLM 5.3 Xplodedglm-5.3-xploded | $2.00 | $1.00 | $5.00 | 1.05M |
| GLM 5.3 Flash Xplodedglm-5.3-flash-xploded | $0.30 | $0.20 | $1.00 | 1.05M |
| DeepSeek V4.1 Flash 1017 Xplodeddeepseek-v4.1-flash-1017-xploded | $1.25 | $0.30 | $2.50 | 1.05M |
| DeepSeek Flash Lite Xplodeddeepseek-v4.1-flash-lite-1017-xploded | $0.36 | $0.03 | $1.44 | 512k |
| MiMo V2.6 Flash Xplodedmimo-v2.6-flash-xploded | $1.00 | $0.50 | $2.00 | 1.05M |
| Qwen3.8 27B Xplodedqwen3.8-27b-xploded | $1.50 | $0.50 | $5.00 | 1.05M |
| Qwen3.8 27B Ultrafast Xplodedqwen3.8-27b-fast-xploded | $1.75 | $0.50 | $5.50 | 524k |
FAQ
Is there a subscription?
Not required. Prepaid credit always works with no seat fee and no monthly minimum: you buy credit and spend it on tokens.
Are the prices on the models page what I am billed?
Yes. The per-model prices are what you pay, so the number you read is the number you are billed.
Do extra API keys cost anything?
No. Sub-keys are unlimited and free. Mint one per engagement, per client or per script and revoke it when the job ends.
What is the minimum top-up?
Five dollars, plus the service fee (3.15% plus 30c). After the first top-up there is no minimum spend.
Is there a service fee?
Yes. Each top-up carries a fee of 3.15% plus 30c, shown before you pay. It goes on top of the credit, so $10 of credit is $10 of credit.
How long are request logs kept?
Ninety days, at no cost. The log holds the model, token counts, cost, latency and outcome. It never holds the prompt or the completion.