Token Factory
Pay for the tokens you use.
No subscription and no minimum spend. Add credit from $10, and each request is paid from it at the model's price.
Text models
Price per one million tokens
In US dollars. Input is what you send, output is what the model writes.
| Model | Model ID | Input / 1M | Output / 1M | Status |
|---|---|---|---|---|
| Qwen-SEA-LION-v4.5-27B-IT | qwen-sea-lion-v4.5-27b |
$0.60 | $3.60 | Live |
| Kimi K3 | k3 |
$2.40 | $12.00 | Live |
| Kimi K2.7 Code | kimi-k2.7-code |
$0.95 | $4.00 | Live |
| Kimi K2.6 | kimi-k2.6 |
$0.95 | $4.00 | Live |
| GLM-5.3 | glm-5.3 |
$1.40 | $4.40 | Live |
| GLM-5.3 Flash | glm-5.3-flash |
$0.10 | $0.35 | Live |
| GLM-5.2 | glm-5.2 |
$1.40 | $4.40 | Live |
| GLM-5.1 | glm-5.1 |
$1.40 | $4.40 | Live |
| GLM-5 Turbo | glm-5-turbo |
$1.20 | $4.00 | Live |
| DeepSeek V4 Pro | deepseek-v4-pro |
$1.32 | $3.96 | Live |
| DeepSeek V4.1 Flash | deepseek-v4.1-flash |
$0.18 | $0.72 | Live |
| MiniMax M3 | minimax-m3 |
$0.30 | $1.20 | Live |
| MiniMax M3.1 Flash | minimax-m3.1-flash |
$0.15 | $0.60 | Live |
| Qwen3.8-27B | qwen3.8-27b |
$0.45 | $3.00 | Live |
| Qwen3.6-35B-A3B | qwen3.6-35b-a3b |
$0.10 | $1.00 | Live |
| Qwen3-30B-A3B-Instruct-2507 | qwen3-30b-a3b-instruct-2507 |
$0.10 | $0.30 | Live |
| MiMo-V2.6-Flash | mimo-v2.6-flash |
$0.14 | $0.28 | Live |
| Gemma 3 27B | gemma-3-27b-it |
$0.10 | $0.30 | Live |
| Hermes 4 405B | hermes-4-405b |
$1.00 | $3.00 | Live |
| Nemotron 3 Ultra 550B-A55B | nemotron-3-ultra-550b-a55b |
$1.00 | $3.00 | Live |
| Nemotron 3.5 Lightning | nemotron-3.5-lightning |
$0.06 | $0.24 | Live |
Reasoning counts as output. Every text model except Qwen-SEA-LION-v4.5-27B-IT, Qwen3-30B-A3B-Instruct-2507, MiMo-V2.6-Flash, Gemma 3 27B and Hermes 4 405B can reason before it answers, so its replies can use more output tokens than their visible text. Kimi K3 does not show its reasoning, but it is counted. See Reasoning and tokens in the docs.
Video models
Price per second of video
In US dollars, for each second of video delivered and each reference image sent with it. Pika H3 Flex also charges each second of reference video it reads, at the rate of the chosen resolution; its reference images and audio are free. A generation that fails is not charged.
| Model | Model ID | Per second | Per reference image | 5-second clip | Status |
|---|---|---|---|---|---|
| MiniMax H3 | minimax-h3 |
$0.15 | $0.05 | $0.75 | Live |
| Pika H3 Flex, 480p | pika-h3-flex |
$0.02 | Free | $0.10 | Live |
| Pika H3 Flex, 768p | pika-h3-flex |
$0.05 | Free | $0.25 | Live |
| Pika H3 Flex, 1080p | pika-h3-flex |
$0.065 | Free | $0.325 | Live |
| Pika H3 Flex, 2K | pika-h3-flex |
$0.07 | Free | $0.35 | Live |
| Pika H3 Flex, 4K | pika-h3-flex |
$0.09 | Free | $0.45 | Live |
How billing works
Prepaid credit, charged per request.
-
Add credit
Buy credit by card through Stripe: $10, $25, $50, $100, or any amount up to $1,000.
-
Pay per request
Each request's input and output tokens are charged at the model's price and taken from your credit.
-
Top up any time
The console warns you when less than $1 is left. A request your credit cannot cover is refused, never billed later.
Questions
Pricing and billing, in detail
Is there a subscription or a minimum spend?
No. You pay for the tokens your requests use, at the price of the model you call. The smallest top-up is $10.
What counts as a token?
Input tokens are everything you send in a request, including system and earlier messages, plus the tokens the model adds itself: its chat format and, for some models, a built-in system prompt. Output tokens are what the model generates, including any reasoning, shown or not. The counts come from the model for each request and appear in your usage; the docs list the models that add the most.
Am I charged for a request that fails?
No. A request that fails before the model returns a response is recorded as not charged.
Can I cap what my workspace spends?
Your credit is the cap: a request is only sent when your credit can cover it. You can also set a monthly spending limit below your credit in the console.
How do I pay?
By card, on a payment page hosted by Stripe. Card details never reach MarsCompute. Your card statement shows MARSCOMPUTE.AI.
Can I see what each request cost?
Yes. Usage lists every request with its token counts and cost, can be filtered by member, key and model, and can be exported as CSV.
Add credit, then send a request.
Sign up and your workspace is ready as soon as you verify your email.