TOKEN FACTORY
Comprehensive AI Compute
Operations Solution
Transform excess GPU capacity into a measurable and profitable token business. Comprehensive coverage of resource management, model serving, and billing operations in one package.
Solution Principles
Drive AI compute with industrial-scale efficiency
Industrialized compute manufacturing
Transform diverse GPU pools into token manufacturing lines with integrated scheduling and quality assurance for massive scale.
Monetizing via tokens
Integrated token tracking, billing, and multi-tenant financial settlement — enabling every GPU to generate tangible income.
Integrated models, traffic, and business
A unified management interface covering model lifecycles, traffic distribution, and tenant administration.
Trackable, verifiable, and scalable
Detailed request logging, service level monitoring, and compliance reviews ensure compute operations remain clear and predictable.
Key Capabilities
Full-stack tools to launch your business today
Automated resource orchestration
Organize multi-vendor GPU resources into streamlined token production units with intelligent load balancing and quality metrics.
Model commercial management
Facilitate the launch of 100+ popular models, utilizing dynamic quantization and performance-based pricing to treat models as products.
Comprehensive billing cycle
Accurate metering across tokens, runtime, and concurrency. Flexible payment structures with automated invoicing and revenue distribution.
Growth via analytics
Real-time dashboards for financial health, user retention alerts, and capacity planning to drive informed growth strategies.
Architecture & Data Flow
End-to-end path from supply to revenue
A cloud-native architecture connecting compute, models, and billing into a highly observable and metered commercial loop.
System Layers
User portal · API collection · SDKs
Multi-tenant isolation with centralized key and quota management.
Smart gateway · policy routing · traffic analytics
Route requests by model context and ensure accurate usage auditing.
Model store · inference nodes · automated evaluation
Versatile model versioning compatible with leading architectures.
Resource pooling · compute virtualization · elastic allocation
Global resource scheduling, dispatching tasks by cost and performance metrics.
Bare metal · multi-cloud · edge compute
Unified management of on-premise and external cloud resources.
Data Flow
- 01
Request Ingress
Clients initiate calls via API; system validates identity.
- 02
Smart Routing
Gateway picks the best path based on cost and latency logic.
- 03
Compute Dispatch
The scheduler assigns instances to fulfill compute tasks.
- 04
Logging & Audit
Capture granular token usage and performance data.
- 05
Settlement & Insights
Aggregate invoices and feed data back to management systems.
Capabilities
Six pillars for a robust operation
Compute asset management
- Supports diverse GPU vendors and accelerators
- Automated routing based on latency and cost
- Fast elasticity and automated fault recovery
Model repository
- Rapid deployment for 100+ open-source models
- Multi-tier caching and quantization optimization
- Standardized evaluation and quality testing
Precision metering & billing
- Multi-mode billing for tokens and duration
- Versatile quota and payment configurations
- Automated billing and profit sharing logic
Tenant operations hub
- Hierarchical control for projects and keys
- Configurable rate limits, budgets, and alerts
- Fully automated user onboarding experience
Operational data hub
- Holistic revenue and expenditure dashboards
- User profiling and retention diagnostics
- Procurement guidance and demand forecasting
SLA & monitoring stack
- SLA monitoring with automated credit issuance
- Export support for standard telemetry tools
- Deep request tracing and performance analysis
Strengths
Multi-dimensional advantages for your business
Swift delivery & agility
- Containerized setup allows for a ~2 week launch window
- Over 100 pre-configured models ready for immediate use
- Intuitive visual interface minimizes manual coding requirements
Performance & yield boost
- Deeply tuned inference logic provides major throughput gains
- Advanced caching and batching for optimal resource use
- Maintains superior GPU utilization under heavy workloads
Cost optimization & savings
- Dynamic quantization greatly reduces compute footprints
- Semantic caching accelerates responses for common queries
- Intelligent dispatching for the most cost-effective path
Unified commercial lifecycle
- Supports diverse billing metrics and business models
- Seamless integration of payments, quotas, and invoicing
- Automated reconciliation and revenue-share workflows
Secure & auditable
- Private cluster deployment options for strict data sovereignty
- Robust tenant isolation and security safeguards
- Compliance frameworks aligned with global security standards
Growth insights & analytics
- Real-time visualization of key business metrics
- Granular user analytics and churn prevention
- Evidence-based capacity forecasting and buying guides
Target Users
Crafted for every compute supplier
Compute center operators
Convert raw compute resources into sellable MaaS offerings, accelerating the journey to profitability.
Industry cloud providers
Offer white-labeled token services integrated with niche models to gain a competitive edge.
Platform companies & SaaS
Establish a centralized AI foundation for internal needs and external monetization via API access.
Use Cases
Real-world token businesses across sectors
Regional compute platforms
Unify various accelerators into standard token services for regional industries as a MaaS hub.
Operator cloud services
Harness edge computing for low-latency inference tailored to specific industry needs.
Research & education cloud
Flexible token allocation for academia, streamlining model tuning and compute scheduling.
Medical & life sciences
Compliant compute power for diagnostic imaging, record processing, and drug discovery.
Commerce & content hubs
High-concurrency APIs for content generation and personalized commerce experiences.
Client Feedback
Helping operators embrace intelligent commerce
「After implementing SmarTokenX, we productized our compute in weeks. Per-unit yield and revenue grew significantly on the same hardware.」
A regional compute center lead
「Token Factory enabled us to build bespoke MaaS for clients. The built-in billing and operations drastically cut our delivery overhead.」
A leading global systems integrator
「This solution unified our base and closed the commercial loop, helping us shift from a cost center to a profit-generating unit.」
An internet giant's AI group
Implementation Path
Four steps from assets to profit
Assess & Research
Evaluate compute assets and goals to design optimal capacity and pricing.
Deploy & Setup
Rapid installation via Helm, integrating resources and setting up billing paths.
Operate & Scale
Acquire users through portals and channels, using real-time insights for optimization.
Optimize & Monetize
Plan expansions based on data, leveraging partnerships to increase profit margins.
Management API Examples
Compute Management APIs with detailed documentation
Token Factory provides operators with a full suite of control APIs covering tenant authorization, quota allocation, and billing. Below are the core interaction specifications.
Method 1: Key Pair
Best for server-side use; supports creation, rotation, and revocation via console.
curl -X POST https://api.your-domain.com/v1/tenants \
-H "X-API-Key: stx_tk_xxxxxxxx" \
-H "X-API-Secret: stx_sk_xxxxxxxx" \
-H "Content-Type: application/json" \
-d '...'Method 2: JWT Token
Ideal for console actions and automation scripts with role-based claims.
curl -X POST https://api.your-domain.com/v1/tenants \
-H "Authorization: Bearer eyJhbGciOiJIUzI1NiIs..." \
-H "Content-Type: application/json" \
-d '...'X-API-Key / X-API-Secret — Operator credentials; obtain from the management console.
Authorization: Bearer <token> — Token from login endpoint; access restricted by role.
Content-Type — Defaults to application/json.
X-Request-ID — Optional; for end-to-end tracing purposes.
Request Body
// Headers:
// Authorization: Bearer <operator_jwt>
// Content-Type: application/json
// X-Request-ID: req_xxx
{
"name": "Acme Corp",
"contact_email": "admin@acme.com",
"plan": "enterprise",
"quota_tokens": 100000000,
"rate_limit_rpm": 10000,
"billing_currency": "CNY",
"tags": ["finance", "internal"]
}name — Tenant name
contact_email — Admin email
plan — Subscription plan type
quota_tokens — Initial quota amount
rate_limit_rpm — Requests per minute limit
billing_currency — Billing currency selection
tags — Tags for organizational grouping
Response Body
{
"tenant_id": "tnt_abc123",
"api_key": "stx_tk_xxxxxxxx",
"api_secret": "stx_sk_xxxxxxxx",
"quota_balance": 100000000,
"status": "active",
"created_at": "2026-06-18T08:00:00Z",
"portal_url": "https://portal.your-domain.com/t/tnt_abc123"
}tenant_id — Unique tenant identifier
api_key / api_secret — Access credentials; store securely
quota_balance — Remaining token balance
status — Current account status
portal_url — Link to the dedicated portal
Query Parameters
// Headers:
// Authorization: Bearer <operator_jwt>
// X-Request-ID: req_xxx
// Query parameters
tenant_id=tnt_abc123
&start=2026-06-01T00:00:00Z
&end=2026-06-18T23:59:59Z
&granularity=day
&group_by=model
tenant_id — ID for single or batch lookups
start / end — Time range in ISO-8601 format
granularity — Time precision for grouping
group_by — Dimension for data aggregation
Response Body
{
"period": { "start": "...", "end": "..." },
"summary": {
"total_tokens_in": 450000000,
"total_tokens_out": 120000000,
"total_cost_cny": 28500.00,
"total_revenue_cny": 52000.00,
"gross_margin_rate": 0.452
},
"items": [
{
"dimension": "deepseek-v3",
"tokens_in": 200000000,
"tokens_out": 50000000,
"cost_cny": 12000.00,
"revenue_cny": 22000.00,
"avg_latency_ms": 245,
"availability": 0.9992
}
]
}total_tokens_in / out — Input/output token totals
total_cost_cny — Estimated resource expenditures
total_revenue_cny — Total accumulated revenue
gross_margin_rate — Profit margin percentage
avg_latency_ms / availability — SLA & performance metrics
Request Body
// Headers:
// Authorization: Bearer <operator_jwt>
// Content-Type: application/json
// X-Request-ID: req_xxx
{
"model_id": "deepseek-v3",
"name": "DeepSeek V3",
"pricing": {
"input_per_1m_tokens_cny": 2.00,
"output_per_1m_tokens_cny": 8.00,
"currency": "CNY"
},
"regions": ["cn-beijing", "cn-shanghai"],
"gpu_spec": "H100-80GB",
"instance_count": 4,
"quantization": "fp8",
"enabled": true
}model_id — System identifier for model
pricing — Pricing per million tokens
regions — Target deployment regions
gpu_spec — Required GPU specifications
instance_count — Initial node count
quantization — Selected quantization level
Response Body
{
"model_id": "deepseek-v3",
"status": "deploying",
"endpoints": [
{
"region": "cn-beijing",
"url": "https://infer.cn-beijing.your-domain.com/v1",
"health": "healthy",
"current_tps": 1280
}
],
"estimated_ready_at": "2026-06-18T10:00:00Z",
"revision": "v1.0.0"
}status — Current deployment stage
endpoints — Service endpoints by region
current_tps — Live throughput statistics
estimated_ready_at — Estimated service ready time
Help & Support
Common Questions
Difference between Token Factory and Enterprise MaaS?
Typical timeline from setup to revenue?
Are white-label options available?
How to calculate expected ROI?
Support for mixed compute resources?
Does it include automated top-ups?
Integration with existing ERP/finance?
Start monetizing your compute today
Share your resource details and business vision — our experts will help you craft a custom launch plan and ROI roadmap.