With dat1, you don't need to manage hardware and think about scaling. We manage everything in the background and reduce operational and hardware costs by efficiently sharing the GPUs between our customers.
You don't need to worry about high fixed costs typically associated with GPU inference, we charge per second ($0.005 per second for an NVIDIA A100) and we only charge for the time your model runs inference—no costs for idle time or timeouts.
We worked hard to reduce cold starts which is an important point in all serverless platforms. With dat1, you can expect a 15-20-second cold start for a 70Gb model. We will also work closely with you to optimize it.
We worked hard to reduce cold starts which is an important point in all serverless platforms. With dat1, you can expect a 15-20-second cold start for a 70Gb model. We will also work closely with you to optimize it.
We value our customers' privacy. We are fully GDPR and CCPA-compliant and follow top cybersecurity best practices.
