Koyeb

Koyeb GPU pricing overview

Koyeb is a specialist GPU cloud provider headquartered in Paris, France. Koyeb is joining Mistral AI to build the future of AI infrastructure. The platform offers serverless GPU compute with automatic scaling to zero and per-second billing. GPU catalog: B200, H200, H100, A100 SXM, RTX PRO 6000, A100, L40S, RTX A6000, L4, RTX 4000 SFF Ada. N300s (NVIDIA Rubin) available via request. 7 regions in US, EU, and Asia. Saving plans up to 50% off. Kubernetes deployment in 120 seconds. Check koyeb.com/pricing for current rates. Billing is Pay-per-use (per-second), Saving plans with a minimum commitment of None. Available regions include EU, US, Asia. GPU instances can be provisioned on-demand, making Koyeb suitable for both short-duration experiments and sustained production workloads.

Koyeb vs other GPU providers

Koyeb competes with providers including Lambda Labs, CoreWeave, RunPod, Paperspace, Vast.ai, and the major hyperscalers (AWS, Google Cloud, Azure) for GPU compute workloads spanning LLM training, fine-tuning, and inference serving. Key differentiators include: B200, H200, H100 — competitive serverless pricing; Joining Mistral AI — strong AI infrastructure future; Automatic scaling to zero — no idle costs. Use the side-by-side comparison tool above to see Koyeb pricing against any other provider across shared GPU models. For a broader market view, the live GPU prices table shows all 2 Koyeb listings alongside 91+ providers in a single sortable view.

Best use cases for Koyeb

Koyeb is best suited for: Inference API deployments with variable traffic, Serverless AI applications, Teams wanting zero-ops GPU inference, Mistral AI ecosystem users. Support tiers range from Community → Enterprise, making it viable for both individual researchers and enterprise teams with SLA requirements. There are currently 2 active GPU listings on Koyeb, covering L4, A100 80GB. H100 SXM5 instances with NVLink interconnect are commonly used for large-scale training workloads. A100 40GB and RTX 4090 instances are frequently used for fine-tuning and inference of smaller models.

Koyeb billing model and cost structure

Koyeb uses Pay-per-use (per-second), Saving plans pricing. Spot pricing is not currently available on this provider — all instances are on-demand. Reserved instance pricing may be available for predictable long-running workloads — check the provider's pricing page for current committed-use rates. Always compare the effective hourly rate including egress, storage, and networking costs when evaluating total cost of ownership across providers.

Choosing the right GPU on Koyeb

GPU selection depends on model size, precision, and whether your workload is compute-bound or memory-bandwidth-bound. For LLM training above 30B parameters, H100 80GB SXM5 instances with NVLink are the standard choice — high HBM3 bandwidth and FP16 throughput make them well-suited for large transformer workloads. For inference of 7B–13B models in FP16 or BF16, A100 40GB is a common choice on many providers. RTX 4090 instances are widely used for fine-tuning, prototyping, and quantized inference (INT4/INT8) of models up to 70B. Read the H100 vs A100 guide or the GPU benchmarks for ML guide for a full breakdown.

How Koyeb pricing data is collected

Prices shown are sourced from Koyeb's public pricing API or pricing page and refreshed every 15 minutes. On-demand rates reflect the current list price for a single GPU instance in the cheapest available region. Spot prices, where available, reflect interruptible instance rates at the time of the last snapshot. All prices are in USD per hour. Daily snapshots are kept in our history archive and visualised in the GPU price history charts — useful for identifying seasonal pricing patterns and evaluating whether current rates are above or below the 30-day average.

Evaluating managed LLM inference APIs as an alternative to self-hosted GPU compute? Compare live LLM token prices across OpenAI, Anthropic, Google, Groq, and many other providers. The cheapest GPU cloud guide covers the break-even analysis between self-hosted and managed inference at different request volumes.