GPU compute
Customer priceLoading live GPU pricing…
Fine-tuning
Estimated per 1M tokensModel SizeSFTPreference/RL
Up to 8B$0.45$1.10
9B-34B$0.90$2.20
35B-70B$1.80$4.40
70B+CustomCustom
Billing is based on GPU-hours, runtime, and storage — you pay for compute time, not per-token markup.
Deployments
GPU-hour basedModeBillingNotes
On-demandGPU-hour tableAutoscale by endpoint capacity
DedicatedContact usReserved GPUs for production traffic
Clusters
CapacityTypeAvailabilityPrice
On-demandAuto-selected providersGPU-hour table
DedicatedReserved capacityContact us
Spot pricing is not exposed as a user-facing option yet. Current jobs auto-select configured providers unless a provider is specified.
Storage
MonthlyItemPriceUnit
Veri Volumes$0.085GiB-month after 512 GiB free
Checkpoint storage$0.023GiB-month
Dedicated serving
Sized to your traffic · compared vs frontier API list pricesCompute CostsInput TPM/GPUCached TPM/GPUOutput TPM/GPUPrice GPU/Min
MiniMax M3138,840694,20023,140—
GLM-5.235,731192,4009,620—
Qwen3 235B52,000260,00011,500—
Comparison
Your traffic on Veri vs frontier API list pricesEstimated num of H100s33
Estimated Monthly CostLoading…
Estimated Monthly SavingsLoading…
Estimates assume continuous 24/7 provisioning (43,800 min/mo) sized to your peak load, billed at an estimated $2.50/GPU-hr (live rate temporarily unavailable). Savings compare against the selected frontier model's published list price (cached input billed at the provider's caching discount) on the same traffic profile. Throughput per GPU varies with context length and serving configuration — talk to us for a sized quote.