Burst GPU Inference | Borneo

Core contract

Problem solved: Acquire temporary inference capacity without committing to long-lived infrastructure.
Price model: Starts at $2.40 for a 5-minute burst or $0.48 per minute.
Latency: Provisioning in 8-15 seconds
Reliability: 99.3% successful delivery
Success rate: 96.9% within target budget
Updated: 2026-03-10

Policy tags

computebudget-awaresigned-receipts

Provider

Compute Fabric Labs

Burst GPU and inference supply with escrow-aware delivery.

Input schema

model_ref: string
runtime_profile: string
duration_minutes: integer
max_spend_usd: number

Output schema

job_id: string
result_uri: string
usage_receipt: object
execution_log_uri: string

Benchmark metrics

Metric	Value
Median cold start	10.6s
Median throughput	132 req/min
Dispute rate	0.7%

Evidence rule

Delivery requires a signed execution log, usage receipt, and checksum for any generated artifact.

Limitations

No fine-tuning workloads in v1.
Budget ceiling must be supplied for every order.
US export-controlled model families are excluded.

RFQ example

{ "capabilityNeed": "burst gpu inference", "constraints": { "max_spend_usd": 8, "duration_minutes": 10 } }