Koyeb is a serverless cloud platform for deploying apps, APIs and GPU AI workloads. See pricing, GPU options and how it compares to Render and Fly.io.
Koyeb is a serverless cloud platform founded in 2020 by three former Scaleway employees, Yann Léger, Édouard Bonlieu and Bastien Chatelard, and headquartered in Paris, France. It lets developers deploy applications, APIs and AI inference workloads directly from Git or Docker without managing servers, with automatic scaling, global distribution and per-second billing.
In February 2026, Mistral AI acquired Koyeb as its first-ever acquisition, bringing the roughly 13-person team into Mistral's engineering organization to accelerate Mistral Compute, the AI lab's cloud infrastructure initiative, signaling Koyeb's evolution from an independent developer platform toward core AI infrastructure for one of Europe's leading AI labs.
Koyeb's core platform deploys apps directly from a Git repository or Docker image and handles autoscaling, load balancing, TLS and global edge distribution automatically, with services able to scale to zero when idle and back up on demand, billed per second of actual usage. This makes it a git-push deployment experience without the operational burden of managing Kubernetes or raw VMs.
For AI and machine learning workloads, Koyeb offers on-demand NVIDIA GPU access spanning RTX 4000 SFF up through A100, H100, H200 and B200 configurations, all billed per second with no minimum commitment, alongside a managed serverless Postgres database product with autoscaling storage for applications that need a bundled data layer.
Koyeb offers a genuinely free Hobby plan (one web service plus a small Postgres instance, no credit card required, never expires) alongside three paid tiers: Pro at $29/month with $10 of compute credit, Scale at $299/month with $100 of compute credit and a 99.9% SLA, and a custom Enterprise plan starting around $1,000/month with $500 of compute credit, unlimited users and a 99.99% SLA.
GPU compute is billed separately and per second, ranging from about $0.50/hour for an RTX 4000 SFF up to $24/hour for an 8x H200 configuration, with committed-capacity discounts of up to 50% available for teams that reserve capacity long-term, and an early-stage startup program offering up to $30,000 in credits.
Yes, Koyeb's Hobby plan is free forever, includes one web service and a small Postgres database instance, requires no credit card, and never expires, though it is not intended for production traffic.
Paid plans start at $29/month for Pro (with $10 compute credit), $299/month for Scale (with $100 compute credit and a 99.9% SLA), and custom Enterprise pricing starting around $1,000/month.
Yes, Koyeb offers on-demand NVIDIA GPUs including RTX 4000, A100, H100, H200 and B200, billed per second starting around $0.50/hour.
Koyeb was acquired by Mistral AI in February 2026, its first acquisition, to accelerate Mistral's cloud infrastructure ambitions.
Koyeb is headquartered in Paris, France, and was founded in 2020 by three former Scaleway employees.
Koyeb bills per second of actual compute usage, including GPU instances, with services able to scale to zero when idle.