Atomic CloudAI infrastructure and controlled data processing.
Contact sales
Infrastructure for global AI builders

AI infrastructure,
built around controlled compute.

Access DeepSeek, GLM, Qwen and Kimi through one commercially competitive API relationship — with requests processed in the Shantou data center and generated results returned overseas.

Request API access How data processing works

One API patternFour model families

Controlled routeShantou processing

Security-mindedReviewed onboarding

curl "$ATOMIC_CLOUD_BASE_URL/chat/completions" \
  -H "Authorization: Bearer $ATOMIC_CLOUD_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "deepseek-v4-flash",
    "messages": [{"role": "user", "content": "Hello"}],
    "stream": true
  }'
OpenAI-style request schemaFinal base URL and model IDs are supplied during onboarding
Assisted API onboarding availableStart a technical review

Launch model portfolio

One API. Four leading model families.

Switch between model strengths without rebuilding your commercial and integration workflow.

Qwen

Qwen3.5-122B-A10B

Multilingual + multimodal

View model
Compare model fit

Unified API

A familiar request format across the launch model portfolio.

Commercially competitive

Volume, cache behavior and model mix inform each proposal.

Controlled data path

Requests route to designated compute in the Shantou data center.

Usage visibility

Token monitoring and account controls are on the product roadmap.

Bring Data to Compute · 来数加工

Move the request to designated compute. Return the result.

Data is created outside China, transmitted by API to the Shantou data center for agreed processing, and the generated output is returned to the overseas application.

Review the processing architecture
CONTROLLED REQUEST PATHOverseas application → Shantou compute → Overseas result
Onboarding review required
01

Overseas application

A purpose-limited request is created in the customer environment.

02

Encrypted API transit

Client-specific credentials and encrypted transport control access.

03

Shantou processing

Model inference runs inside the agreed processing environment.

04

Result returned

The generated output is delivered back to the overseas application.

Encrypted transport Scoped credentials Purpose-defined processing Contracted controls

Compliance-by-design, not an automatic exemption. Legal treatment depends on the actual data, logs, retention, contracts, technical configuration and applicable jurisdictions. Specific claims apply only when confirmed in writing.

Platform roadmap

A control plane built for what comes next.

Token monitoring, customer accounts and payment workflows are planned as the next layer after assisted API access.

OverviewProduct interface preview
ROADMAP PREVIEW
Token usageUsage visibilityPlanned
API requestsRequest analyticsPlanned
Model routes4 launch familiesAvailable
Account fundingPayment workflowPlanned
Token usageIllustrative interface
EarlierLater
Request volumeIllustrative interface
EarlierLater
Cost visibilityIllustrative interface
EarlierLater

Commercial model

Transparent quotes built around real workloads.

Pricing is prepared using model choice, cache behavior, input/output mix, target region and committed usage.

View pricing approach
ModelCore fitInput structureBilling structureAccess
DeepSeek V4 FlashFast text + codeCached / uncachedInput + outputRequest quote
GLM-5.1Agentic engineeringCached / uncachedInput + outputRequest quote
Qwen3.5-122B-A10BMultilingual + multimodalCached / uncachedInput + outputRequest quote
Kimi K2.6Reasoning + tool useCached / uncachedInput + outputRequest quote
No public rate is represented as final until confirmed in a quotation.AI inference tokens — not cryptocurrency.
curl "$ATOMIC_CLOUD_BASE_URL/chat/completions" \
  -H "Authorization: Bearer $ATOMIC_CLOUD_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "deepseek-v4-flash",
    "messages": [{"role": "user", "content": "Hello"}],
    "stream": true
  }'
OpenAI-style request schemaFinal base URL and model IDs are supplied during onboarding

Built for developers

Familiar integration. Direct implementation support.

Start with an OpenAI-style request pattern. Exact endpoint details, production model identifiers and credentials are provided during onboarding.

OpenAI-style schemaKeep the request pattern familiar.

Streaming structureSupport responsive product experiences.

Model portfolioEvaluate four leading model families.

Scoped credentialsSeparate access by customer and environment.

Usage reconciliationAlign commercial records with usage.

Direct supportOne contact for technical and commercial questions.

Request technical onboarding

Need billions of tokens?

Enterprise AI Infrastructure

Volume-aware pricing, direct implementation support and tailored API onboarding for high-volume AI workloads.

Contact sales Learn more
  • Volume-aware pricing
  • Direct technical support
  • Processing-path review
  • Data handling & security
  • Region planning