Back to models

DOS.AI Auto analyzes each request and classifies it as Simple, Medium, or Complex. Simple and Medium requests use DOS.AI, while Complex requests can use an eligible paid route configured in the live catalog. The OpenAI-compatible endpoint keeps one integration, and response headers identify the requested and provider model for inspection.

Capabilities:Automatic model selectionCost optimizationTool callingStreamingOpenAI-compatible
Modalities
Input / Output Price
Usage-based
Context Length
128K tokens
Parameters
-

Smart Router Architecture

Auto-Tiered Engine

DOS Smart Router (model: "auto") dynamically classifies query intent, token volume, and reasoning depth in < 4ms to dispatch across 3 destination tiers with zero markup and 99.99% circuit breaker auto-failover.

Selected Models/Top Models Used by DOS Router4 Active Targets

Real-time dispatch distribution across specialized destination models based on prompt complexity and context requirements.

1.
DOS (Qwen 3.8 27B Dense)Tier 0 / 1: Fast & Agent Pre-Tuned
by DOS.AI•262K context
582.4B
59%
2.
GPT-6 LunaTier 1: Balanced Reasoning
by OpenAI•1M context
18.4B
1.9%
3.
Gemini 3.8 FlashTier 0: Sub-second TTFT
by Google•1M context
248.2B
25.1%
4.
Qwen 3.8 MaxTier 2: High Reasoning
by Alibaba Cloud•1M context
139.1B
14%

Destination Tiers

Tier 0
Fast & Efficient

Simple tasks, quick answers.

Latency:< 200ms TTFT
Workloads:Chit-chat, summaries, translation
Targets:DOS Fast, Qwen 2.5 7B
Tier 1
Balanced

Everyday tasks, balanced capability.

Throughput:High tokens/sec
Workloads:Code generation, JSON, agents
Targets:DeepSeek V3, Claude 3.5 Haiku
Tier 2
Advanced Reasoning

Complex problems, deeper reasoning.

Intelligence:Frontier SOTA Reasoning
Workloads:Deep CoT, complex math, architecture
Targets:Claude 3.7 Sonnet, DeepSeek R1

5-Step Routing Lifecycle

01model: "auto"
Request

Your application sends a standard OpenAI-compatible request with model set to auto.

02Simple • Medium • Complex
Complexity Check

DOS classifies the request as Simple, Medium, or Complex.

LengthCodeReasoningTechnical
03DOS.AI • Eligible Route
Route Decision

Simple and Medium use DOS.AI. Complex can use an eligible route from the live catalog.

04Catalog-driven
Provider Request

The gateway sends the request through the route resolved from the tier and current catalog configuration.

99.99% Auto-Failover
05X-Provider-Model
Visible Route

Response headers identify the requested and provider model so your application can inspect the route.

X-Provider-Model
Response Header Transparency Inspector
HTTP/1.1 200 OK
// Upstream audit metadata returned in every response
x-provider: "anthropic"
x-provider-model: "anthropic/claude-3-7-sonnet"
x-dos-route-tier: "tier-2-reasoning"
x-dos-routing-latency: "3.2ms"
x-dos-retail-markup: "$0.00 (0% Zero Markup)"
x-dos-failover: "none (primary upstream healthy)"
Respect Project BYOK

Honors your custom provider credentials with zero platform markups.

99.99% Circuit Breaker

Instant multi-provider failover when upstream encounters 429s or outages.

Zero-Markup Policy

100% official publisher list price. All enterprise discounts belong to internal margin.

Hard Budget Limits

Enforces strict per-key and agent spending ceilings before request dispatch.

API Integration

Use the DOS API to integrate Auto into your applications. Compatible with standard OpenAI SDKs for easy drop-in migration.

</>Run inference
curl https://api.dos.ai/v1/chat/completions \
  -H "Authorization: Bearer $DOS_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "auto",
    "messages": [
      {
        "role": "user",
        "content": "Hello!"
      }
    ]
  }'