Product Plan
Run your product’s AI on a monthly plan
A monthly base allocation for your product, agent, or service, licensed for the users of your own product. Your order defines the allocation and service terms.
Choose one model or all three.
Each model has a separately specified base allocation shared across your product’s concurrent requests; your signed order defines the model, allocation, measurement, and any commitment.
80 tokens/s
135 tokens/s
550 tokens/s
Your order defines the model, base allocation, measurement conditions, and any service commitments. Listed model specifications describe the current lineup and may change under the Terms; they do not promise fixed model versions indefinitely. These figures are not a single request’s speed or a monthly token allowance. As a planning figure, 135 output tokens per second is about 1,200 answers of 400 tokens an hour at full use.
Keep your base.
Add capacity when you need it.
Your base stays available during busy periods under the conditions in your order. Free extra capacity is available while the shared pool has room. Auto scaling requires your authorization and available capacity.
| Plan component | Price | What it covers |
|---|---|---|
| One model allocation | $1,200 per unit, per model, per month | Kimi K3, GLM-5.3, or DeepSeek V4.1 Flash. Includes that model’s base allocation. Add units of the same model to scale that base in $1,200 steps. |
| Three-model bundle | $3,000 per month | One allocation each of Kimi K3, GLM-5.3, and DeepSeek V4.1 Flash. Save $600 per month compared with three individual plans. |
| Free extra capacity | No extra charge | Capacity above your base only while the shared pool has room. It is not reserved. |
| Auto scaling | $5 per extra unit, per two-hour period | Auto scaling adds capacity for traffic peaks after your authorization, subject to availability and plan terms. |
One unit is that model’s base allocation shown above. Billing is monthly with a one-month minimum; adding units mid-month charges only the prorated difference for the rest of the month.
Reserved capacity, sold as a monthly plan.
Inference is usually sold per token or by the hour, and a Product Plan unit is that
capacity at a published monthly price.
| How you buy | What you pay for | Where it fits |
|---|---|---|
| Pay per token (Standard API) | Input and output tokens as you use them | Variable or early traffic |
| Dedicated GPUs by the hour | A GPU, whether or not it is busy | Teams that run their own serving stack |
| Provisioned throughput by the hour | A throughput unit metered per hour, usually on a reservation | Enterprise contracts sized in tokens per minute |
| Product Plan | One model’s output throughput for a whole month at a published price, plus $5 per extra unit, per two-hour period | A product whose AI runs all month with peaks |
Connect your product.
Workload, allocation, price, and integration details are confirmed before paid use.
Share the workload
Include normal and peak traffic, typical input and output lengths, and required response behavior.
Review the plan
Review the model, base capacity, auto scaling charges, data terms, renewal, and cancellation.
Prepare the integration
Agree on authentication, environments, monitoring, and the activation process.
A data platform for your product’s data is coming soon.
A Product Plan data platform is planned and is not released. Availability, scope, pricing, storage terms, and controls will be announced separately.
Your data is for your use only, and storage is optional.
Request content
Product Plan inference is Zero Data Retention (ZDR): Inputs and Outputs are not kept as content logs, histories, or backups. Temporary processing and caching follow the Terms, §5.1.
Data control
The planned data platform is not released; any availability, scope, pricing, storage terms, and controls will be announced separately.
Operational metadata
Content-free service metadata is identifiable for up to 90 days from the original event; retained data after that is sufficiently de-identified. Account, billing, tax, and support records follow separate criteria.
Model training
We do not use content from ZDR-covered inference to train models, including where a customer separately opts in. Optional storage and any future separate training submission have their own terms and do not remove inference ZDR.
Questions about Product Plans.
Choose by the workload and service conditions you need.
How is this different from API?
Both power AI features in your product. API bills input and output tokens. Product Plan includes monthly base capacity with no monthly token cap, free extra capacity when available, and optional auto scaling. A data platform for your product’s data is coming to Product Plan but is not released yet. The Product Plan capacity calculator sizes a plan from your traffic and prices the same month per token. Compare API rates
Will each answer arrive faster?
Response time depends on the model, request size, concurrency, and service conditions. Review the defined base commitment and any latency commitment.
Can my product’s users send requests through it?
Yes. Product Plan covers the AI features inside your own product, and the Terms allow integration into products you make available to your own users. Your application manages users, permissions, and the customer experience.
Who manages the end user?
Your application manages users, permissions, and the customer experience. The integration plan defines service responsibilities.
Can I resell the allocation?
You may use a Product Plan for the AI features inside your own product, including serving your product’s users; the Terms allow integration into products you make available to your own users. Resale of the allocation itself is permitted only where your order and the selected model’s license allow it.