Website terms.
This website presents Seb Labs research, product demonstrations, and commercial offerings. Credit purchases and subscriptions use Stripe-hosted Checkout. The connected account may operate in sandbox mode, which is identified at checkout and does not collect a real payment. Enterprise work is scoped through onboarding before a paid engagement.
Credits and billing
Each verified Google identity receives $5 in starter credit once. Launch API pricing is $0.021 per million model input tokens, including prompt templates and question text; output is free. State is processed and metered for each question. Failed inference is not charged. Credits have no cash value and roll over while the account remains in good standing.
One-time purchases add the dollar amount purchased. Optional monthly plans cost $10, $25 or $100 and add $12, $30 or $125 in API credit for each paid monthly invoice. Cancel future renewals in Manage billing; existing credits remain available. Checkout states the price and renewal terms before payment.
Automatic top-ups are off by default. Enabling them authorizes charges to your saved card for your selected amount whenever usage takes your balance below your threshold, subject to your monthly automatic-charge cap. Saving enabled settings while already below the threshold can charge immediately. Disable them anytime to stop future attempts; a payment already initiated may complete. Manual purchases and subscription renewals are separate from that cap. Refunds, chargebacks and disputes revoke related credits; any already-spent refunded credit may leave a negative balance. Billing questions use the merchant contact information on your Stripe receipt.
Quant VPC and custom weight prices are starting points. An onboarding request creates a draft quote for review, without collecting payment or promising immediate infrastructure delivery. Scope and delivery terms must be agreed before an enterprise subscription begins.
Demonstrations and model outputs
Preset distributions are illustrative examples. Custom inference is available only when an authenticated model backend is configured. Model outputs can be wrong and require evaluation for your use case. The website is provided without a service-level commitment.
Research measurements
Benchmarks reflect the datasets, model versions, hardware, and evaluation conditions in the supplied materials. They do not guarantee future performance. Simulated robotics results do not establish physical safety. Financial backtests are historical research and are not investment advice, a recommendation to trade, or a promise of returns.
Responsible use
Use the playground only with inputs you are authorized to process. Do not attempt to bypass authentication, disrupt the service, or submit unlawful content. Before deploying in a consequential environment, validate the model and retain the controls required for that environment.
Open weights and brand assets
Model weights are governed by the license published in the model repository. Downloadable brand and pitch materials retain their original ownership and source context. Research roadmap items are not service guarantees. Checkout prices and an accepted enterprise scope govern any paid service.