PRICING & INFRASTRUCTURE
Frontier AI Research &Lab Infrastructure
High-throughput model endpoints, autonomous research loops, and physical AI compute.
+
+
Compute & Model API
Pay-per-token model endpoints, verified kernels, and controlled inference harness.
Pay As You GoTop-up from $5 • Zero seat fees
Autonomous Research Lab
SFT/DPO/RL training orchestrator, 15-point benchmark suite, and rollout stores.
From $190 / moIncludes monthly research compute
Physical AI & Clusters
PROBE-1 (Humanoids), AEROPROBE-1 (Aerospace), and dedicated H100 bare-metal clusters.
Custom ContractsPrivate VPC & Air-Gapped Deployments
Pay as you go
# No subscriptionHosted Model APIs, per token, on one prepaid balance.
$
$5 minimum top-up. Pay per token.~333.3M tokens estimate
What's included
OpenAI-compatible chat completions and Anthropic-compatible Messages
Per-token pricing with cached input discounts on supported models
Session-aware routing keeps agent replays warm
One prepaid balance, no seats, no per-user fees
Get $1 free to test the Model APIs.
Direct Signup & Add CreditsEnterprise
Dedicated infrastructure, compliance, and custom volume.
Custom pricing
Everything self-serve includes, plus
- Audit logs and role-based access control (RBAC)
- Custom volume and contract terms
- Custom SLAs up to 99.99%
- Enterprise compliance and dedicated security controls
- Dedicated CSM and private Slack
MODEL APIS
Pay for model tokens, not idle capacity
Model
Input
Cached input
Output
PROBE-1
$0.30
$0.0398.2% cache hit, last 24 hours
$1.20
View model
AEROPROBE-1
$0.40
$0.0496.5% cache hit, last 24 hours
$1.60
View model
DeepSeek V4.1 Flash 1017
$0.15
$0.0199.5% cache hit, last 24 hours
$0.60
View model
GLM 5.3
$1.30
$0.1399.1% cache hit, last 24 hours
$4.20
View model
GLM 5.3 Flash
$0.11
$0.0399.3% cache hit, last 24 hours
$0.45
View model
Nemotron 3.5 Lightning 30B
$0.05
$0.01
$0.15
View model
Qwen3.8 27B
$0.10
$0.0195.7% cache hit, last 24 hours
$0.40
View model
Ornith 1.5 35B
$0.10
$0.0194.3% cache hit, last 24 hours
$0.40
View model
DeepSeek V4.1 Flash
$0.14
$0.0398.8% cache hit, last 24 hours
$0.58
View model
Limits and modality-specific capabilities are listed on each model page.
Compare Architecture Tiers
Self-serve compute is funded by flexible top-ups. Lab and Physical tiers add autonomous orchestration and private cluster hardware.
Feature
Compute (PAYG)
Autonomous Lab
Physical AI Clusters
Frontier Model APIs
All open-weights & Horos models
All models + fine-tuned adapters
Custom checkpoints + private weights
Inference Engine
vLLM, SGLang & TensorRT-LLM
Auto-tuned kernels & Pareto search
Dedicated bare-metal H100/H200 nodes
Prefix & State Caching
Up to 90% discount on cache hits
Warm trajectory & rollout memory
Zero-latency dedicated NVMe KV cache
Autonomous Research Loops
SFT, DPO, GRPO & RFT pipelines
Closed-loop hardware-in-the-loop (HIL)
Physical AI Programs
API access only
MuJoCo & Isaac Sim rollouts
PROBE-1 & AEROPROBE-1 private deployments
Outcome Verification
Token & latency traces
15-point benchmark harness
Certified p-CBF safety (0.0% damage)
Deployment Target
Multi-tenant cloud API
Isolated team workspace
Private VPC, on-prem & air-gapped
Support & SLA
Community & email
Priority research engineer
99.99% contractual SLA + direct Slack