December 2024 Summaries

December marked significant advancements in serverless technology, introducing several features aimed at enhancing deployment efficiency and cost-effectiveness. Key updates include the Scale-to-Zero feature, which minimizes costs by scaling infrastructure down to zero during periods of no traffic, and a reduction in prices for accessing high-performance serverless GPUs, making AI development more accessible. The One-Click Catalog now allows users to deploy open-source AI models with ease, providing dedicated endpoints and automatic scaling. Additionally, new Pro and Scale plans were launched to accommodate various service management and scalability needs, ensuring tailored resources and pricing. These innovations emphasize improved performance, reduced costs, and simplified workflows, supported by a revamped website reflecting the company's commitment to high-performance infrastructure. The company is also engaging with users for feedback on future features and has announced open job positions for those interested in contributing to the serverless future.

Koyeb has introduced two new plans, Pro and Scale, as part of its pricing strategy for 2025, aimed at providing a progressive path for scaling on the platform. These plans replace the previous Startup plan and offer more users, services, and included compute resources, maintaining a flat fee starting at $29 per month with usage charged per second. The new pricing tiers cater to different needs, with the Starter plan for individuals, Pro for professional teams, Scale for scaling companies, and Enterprise for mission-critical deployments, each with varying levels of compute power, number of users, and services. All plans support inference workloads, APIs, web apps, and databases, with options for CPU and GPU access, autoscaling, and Scale-to-Zero capabilities across multiple global regions. Users can upgrade or downgrade plans easily without any commitments, and Koyeb provides a free tier for those starting on the platform. The company emphasizes clarity and seamless evolution for developers and businesses, and offers support through community platforms and direct contact options for any inquiries.

Koyeb has announced significant price reductions for its serverless GPU offerings, including the L4, L40S, and A100 models, making them more affordable and efficient for AI workloads by combining scale-to-zero, autoscaling, and per-second billing. The revised pricing structure now displays hourly rates for easier readability, although charges remain calculated per second. Users can leverage these cost-effective serverless GPUs to run larger models, enhance prediction capabilities, and improve performance without managing complex infrastructure. Koyeb provides an accessible deployment process via CLI or Dashboard, supporting pre-built containers or direct GitHub integration for application builds. The platform offers multi-GPU configurations, including up to 8x A100 instances, to accommodate a range of model sizes and AI applications. Additionally, Koyeb's expanded AI toolkit includes features like Volumes and enhanced deployment logs, further simplifying the development and scaling of AI projects on their infrastructure.

Scale-to-Zero is now in public preview, offering a serverless infrastructure solution that automatically adjusts workloads on GPU and CPU based on traffic demands, enhancing cost efficiency and resource management. This feature, combined with autoscaling, enables applications to "sleep" and "wake" based on incoming requests, thus optimizing infrastructure for compute-intensive tasks like inference and multi-tenant SaaS deployments. It provides significant cost savings by billing per second of usage, with no charges when services are inactive. Scale-to-Zero supports global deployments without incurring additional fees for extra regions and addresses cold start concerns with a startup time of 1 to 5 seconds, which is expected to improve further. Users can easily configure the Scale-to-Zero feature through a control panel or CLI, setting services to scale down to zero when inactive and automatically scaling up when demand increases, making it a flexible, controllable, and globally customizable solution for managing serverless applications.