Request
Your application sends a standard OpenAI-compatible request with model set to dos-auto.
model: "dos-auto"Build AI Apps. Faster.
Access powerful AI models through a simple API with low-latency inference and smart model routing.
BYOK
Bring your own keys
47
Models, one API
8
Model providers
1,855tok/s
Measured throughput
Model and provider counts are read live from the catalog. Throughput measured August 11, 2026 at concurrency 64.
Trusted by
One Router For Every Prompt.
Automatic by default. Observable when you need control.
Task Analysis
DOS evaluates prompt intent, syntax, token length, and reasoning complexity.
Intent: Code • Reason
Profile Selection
The router maps the task to an optimal profile: Fast, Balanced, or Frontier.
Fast • MoE • Frontier
Model Hit
The request is dispatched to the optimal candidate model (DeepSeek, Gemini, Claude, etc.).
DeepSeek → Gemini
Visible Response
Actual routed model is returned in headers, keeping full observability and budget guardrails.
✓ Auditable Guardrails
Run The Best AI Models With A Single Line Of Code
Access leading open-source and proprietary models for chat, images, code, and more - with smart routing built in.
CHATCODEVISION
D
DOS.AI Auto
Smart Router|128K ctx
Auto-priced
CHATCODE
D
DOS.AI
35B MoE (3B active)|128K ctx
$0.07 / 1M tokens
CHATCODEVISION
Gemini 2.0 Flash
Flash|1M ctx
$0.11 / 1M tokens
CHATVISION
Gemini 2.0 Flash Thinking
Flash Thinking|1M ctx
$0.11 / 1M tokens
CHATCODE
DeepSeek V4 Flash
MoE (fast tier)|1M ctx
$0.15 / 1M tokens
CHATCODEVISION
Gemini 3.7 Flash
Flash (Hybrid)|1M ctx
$0.16 / 1M tokens
Why use DOS?
Build faster, scale easier, and focus on what matters - your product.
See a dated, reproducible benchmark snapshot with its exact workload and measured results.
Benchmark snapshotAugust 11, 2026
Time to First Token
1.069sec
Output throughput
1,855tok/s
Failed requests
0 / 600
Measured workload
dos-ai
1,024 input / 256 output tokens
concurrency 64
Snapshot, not a guarantee. Results vary by prompt, load, route, and model.
Get started today
Start building with DOS in minutes. No credit card required. Get $5 in free credits to explore our API.
Start building for freeFrequently asked questions
Can’t find what you’re looking for? Reach out to our support team at support@dos.ai.
What models does DOS support?
DOS provides access to leading open-source and proprietary models including Llama 4, DeepSeek, Qwen, Gemma, Mistral, and more. Use dos-auto to classify each request and select an eligible route automatically.
How does pricing work?
We use pay-as-you-go pricing based on metered usage. There is no minimum commitment, and current rates are published by model.
Is there a free tier?
Yes. New users get $5 in free credits to explore the API before adding funds.
How fast is the inference?
Performance depends on the selected route, model, prompt, and load. Our public benchmark snapshot includes its workload and measured results.
Which AI models can I access?
You get a wide range of models through one API - our own DOS.AI model plus models from OpenAI, Anthropic, Google, DeepSeek, Qwen, and more. Use dos-auto to classify each request and select an eligible route automatically.
Is my data secure?
DOS uses TLS in transit, scoped API access, and managed secret references. Requests may be processed by the model provider selected for the route.
Do you offer enterprise plans?
Contact our team to discuss dedicated resources, support, and deployment requirements.
What SDKs do you provide?
Our REST API is compatible with OpenAI client libraries. The documentation includes a quickstart and API reference.
How do I get support?
Use our documentation and community Discord, or contact support@dos.ai for account-specific help.