AI model router: one id, the right model per request
An AI router on a prepaid key: send model "auto" and the gateway picks a fast, balanced or frontier model for that request, billed at the model's own price. No router fee, no subscription; pay as you go by card or in USDT, USDC, BTC, ETH, SOL.
curl https://api.relayswitch.io/v1/chat/completions -H "Authorization: Bearer sk-relay-..." -H "Content-Type: application/json" -d '{"model": "auto", "messages": [{"role": "user", "content": "Summarize this thread."}]}'How the AI router picks a model
Three tiers. The request is classified on the API, with no extra model call, from what it carries.
Everything the two tiers below do not claim: short chat turns, classification, extraction, rewrites.
5 models today. Input $0.0582 to $0.09, output $0.1455 to $0.36 per million tokens.
Tools, code blocks, images, a json_schema answer, or a prompt over 16,000 tokens.
5 models today. Input $0.3178 to $1.20, output $0.9987 to $4.50 per million tokens.
A reasoning ask, a high reasoning effort, or tools together with code.
5 models today. Input $1.20 to $6.00, output $7.20 to $30.00 per million tokens.
| Tier | Model | Input | Output |
|---|---|---|---|
| fast | Claude Haiku 5.5 | $0.0582 | $0.1455 |
| fast | GLM 5.3 Flash | $0.09 | $0.15 |
| fast | GPT-6 Luna | $0.0625 | $0.20 |
| fast | MiniMax M3 | $0.09 | $0.36 |
| fast | DeepSeek V4.1 Flash | $0.06 | $0.18 |
| balanced | Claude Sonnet 5.5 | $1.20 | $3.00 |
| balanced | GPT-6.1 Sol | $0.582 | $2.91 |
| balanced | Gemini 4 Argon | $0.60 | $3.00 |
| balanced | Kimi K3 | $0.90 | $4.50 |
| balanced | GLM 5.3 | $0.3178 | $0.9987 |
| frontier | Claude Opus 5.5 | $2.80 | $10.00 |
| frontier | GPT-6 Astra | $3.00 | $15.00 |
| frontier | Claude Fable 5.1 | $6.00 | $30.00 |
| frontier | Gemini 3.1 Pro Preview | $1.20 | $7.20 |
| frontier | Claude Opus 5 | $1.80 | $7.50 |
USD per million tokens, read from the catalog at render. Within a tier, models are tried in this order.
The same router, one catalog tag
A profile picks only models carrying that tag. Send the id as the model, or set it on the key.
auto/uncensoredFiction, character chat and research that gets refused elsewhere.
Only models carrying the uncensored tag in the catalog.
Models today: GLM-5.3 Flash Uncensored, GLM-5.3 Uncensored, DeepSeek V4 Pro Uncensored, Jailbreak-Free Assistant, DeepSeek V4.1 Flash 1017 Uncensored, MiMo V2.6 Pro Uncensored, Uncensored Coding Agent, Nemotron 3 Ultra 550B Uncensored, Kimi K3.1 Pro Uncensored, MiMo V2.6 Flash Uncensored Thinking, Kimi K3 Uncensored
auto/abliteratedOpen-weight builds with the refusal direction removed from the weights.
Only models carrying the abliterated tag.
Models today: DeepSeek V4.1 Flash Abliterated, GLM-5.3 Flash Abliterated, Apex Flash 1 Abliterated, DeepSeek V4.1 Flash 1017 Heretic, GLM-5.3 Flash Apex Abliterated, Abliterated Large Flash, MiMo V2.6 Flash Abliterated, GLM-5.3 Abliterated, Abliterated Model Large, Abliterated Model
auto/securityExploit analysis, malware triage, CTF and red-team work. Alias: auto/cyber.
Only models carrying the security tag.
Models today: Gemma 4 26B A4B Cybersecurity, GLM-5.3 Flash Cybersecurity, Qwen3.8 27B Cybersecurity, Red Team (Flash), OrcaCyber MiniMax M3.1 Flash, OrcaCyber Qwen3.8 Flash Next, Red Team (Pro), CyberHeabsy, OrcaCyber Zero 1.5
auto/roleplayLong character chats in SillyTavern, JanitorAI, Chub and the other front ends.
Only models carrying the roleplay tag.
Models today: Cydonia 24B v4.3, Magidonia 24B v4.3, Gemma 4 31B Novelist, Uncensored AI Roleplay, Qwen3.8 27B Synthia, Artemis v1.1, Skyfall 36B v2, Wayfarer Large 70B, Qwen3.8 27B Fable, Magnum v4 72B
Route a client that cannot change its model name
Set the router on the key instead of in the request.
One select on the key
Create key in the console has an Auto router select: off, auto, or one of the four profiles. A key set to one sends every chat request through the router whatever model the client names.
Works in fixed-model clients
Claude Code, Cursor, Cline, SillyTavern and any other client with a single model string get the tier that fits each request, with no change to the client.
Same balance, same key
The key draws on the prepaid balance like any other, topped up by card or in USDT, USDC, BTC, ETH, SOL. Existing keys get the same select in their settings.
Three lines of difference
Any OpenAI-compatible client, the Messages, Responses or Gemini endpoints: the model id is "auto" and the rest of the request is yours.
curl, with a tier cap
curl https://api.relayswitch.io/v1/chat/completions \ -H "Authorization: Bearer sk-relay-..." \ -H "Content-Type: application/json" \ -d '{ "model": "auto", "messages": [{"role": "user", "content": "Write the migration plan for this schema."}], "auto": {"tier_max": "balanced"} }'# The answer's "model" names the model that replied; x-uplink-auto-tier names the tier.Python OpenAI SDK, with a sticky session
from openai import OpenAIimport os client = OpenAI( base_url="https://api.relayswitch.io/v1", api_key=os.environ["RELAYSWITCH_API_KEY"],) reply = client.chat.completions.create( model="auto", messages=[{"role": "user", "content": "Refactor this function to async."}], extra_headers={"x-session-id": "ticket-4821"}, # same model for 30 minutes)print(reply.model) # the model the router chose; billed at its own priceAI model router: questions
What it picks, what it costs, and how to steer it.
One id, every model, pay as you go
A key set to the router from $10 by card or in USDT, USDC, BTC, ETH, SOL. No router fee, no subscription.