NewClaude Haiku 5.5

Every client.
Every model.
One key.

One OpenAI-compatible API. Pay with crypto
or bring your own provider keys.

Or try chat first
Published models
281
Pay with MINI
+5% credits
Minimum deposit
$0.50
Account or KYC
Not required

More models. One API.

Explore the published catalog without adding another integration.

Browse models

A route around failures.

Automatic provider failover, with the serving route on your response.

How routing works

Crypto in. Credit ready.

Top up with USDC, SOL, or MINI for 5% extra. Pay as you go.

Explore payments

A key is all you need.

Start without an account or KYC. Save your key and dashboard link.

How access works

Find your balance.

See how model effort changes intelligence and cost.

Compare models
GPT-6.1 SolClaude Opus 5.5

Intelligence

Model intelligence against benchmark task costGPT-6.1 Sol: Low effort, intelligence 42.1, $0.1308 per task; Medium effort, intelligence 47.8, $0.2137 per task; High effort, intelligence 50.2, $0.3191 per task; Xhigh effort, intelligence 51, $0.3929 per task; Max effort, intelligence 51.8, $0.7242 per task. Claude Opus 5.5 (Default Fallback): Low effort, intelligence 42.3, $0.5512 per task; Medium effort, intelligence 51.2, $1.336 per task; High effort, intelligence 53.6, $1.8225 per task; Xhigh effort, intelligence 56, $3.4591 per task; Max effort, intelligence 57.6, $5.982 per task405060$0.1$1$10GPT-6.1 Sol, Low: 42.1 intelligence, $0.1308 per taskGPT-6.1 Sol, Medium: 47.8 intelligence, $0.2137 per taskGPT-6.1 Sol, High: 50.2 intelligence, $0.3191 per taskGPT-6.1 Sol, Xhigh: 51 intelligence, $0.3929 per taskGPT-6.1 Sol, Max: 51.8 intelligence, $0.7242 per taskClaude Opus 5.5 (Default Fallback), Low: 42.3 intelligence, $0.5512 per taskClaude Opus 5.5 (Default Fallback), Medium: 51.2 intelligence, $1.336 per taskClaude Opus 5.5 (Default Fallback), High: 53.6 intelligence, $1.8225 per taskClaude Opus 5.5 (Default Fallback), Xhigh: 56 intelligence, $3.4591 per taskClaude Opus 5.5 (Default Fallback), Max: 57.6 intelligence, $5.982 per task

Benchmark task cost (USD · log scale)

Artificial Analysis · 7 Oct 2026

Already at home in your tools.

Keep your workflow. Connect a different model.

All 17 integrations

A familiar setup.
A lot more choice.

Get a MiniRouter key, then add credit or connect your provider keys. Use the same endpoint in your client.

agent-settings.env
OPENAI_BASE_URL=https://api.minirouter.sh/v1OPENAI_API_KEY=mr-live-…OPENAI_MODEL=meta/llama-3.3-70b

Base URLAPI keyModel ID

Agent setup
OpenAI-compatible endpointhttps://api.minirouter.sh/v1

Your providers.
Your keys.

Connect OpenAI, Anthropic or DeepSeek. Same endpoint. Your priority and spend limits.

0% fee with one key per providerTwo or more enabled keys for a provider: 3% of published-rate usage. Your provider bills inference separately. Providers without an enabled key keep using MiniRouter credits.

Make a good setup
your default.

Save a model, instructions, tools, and supported routing settings as a versioned preset. Reuse it across your clients.

Explore presets
Your saved request
model: "@preset/product-writer"
instructions: saved
tools: saved
version: active

Request fields can override saved settings.
Account and key limits still apply.

A few things to know.

Credits buy inference and cannot be withdrawn.

Prompts, responses, and tool content are captured by default in private debug traces, retained for 30 days.

Read the docs

AI-assisted answers about MiniRouter. Leave out keys and secrets.

Always moving forward.

All updates

One key. Your next possibility.

No account. No KYC. Start from $0.50.

Get a key