# Live AI model marketplace — prices and hosts | Inpherio

Source: https://inpherio.co.uk/marketplace

Live

# Every model on the grid, live.

Each one is served by a real machine, at a price its host sets — not us. Compare price, speed and availability, then copy the model ID straight into your code.

Models live—

Text from—

Embeddings from—

Speech from—

**Filters**

Sort by

Can do

Availability

Layout

Loading live data…

Connecting…

How to read the board

#### Prices

Shown the way they're billed, always in **pounds**. A price under a penny shows its real decimals — **£0.004** is less than half a penny — and under each price is what one typical request costs. Hover a price to see what £1 buys. **Text**: per million tokens, one price for what you send, one for what comes back. **Embeddings**: per million tokens sent — the output is free. **Speech**: per hour of audio — the length of the speech made or the recording transcribed, to the second (1-second minimum), not how long the host takes.

#### “From” prices

When several hosts run the same model at different rates you'll see the lowest. Requests go to the best-scoring healthy host — price is 20% of the score, alongside whether the model is loaded, reliability and speed — so a request can go to a host above the lowest price. Set a maximum price in your preferences (or per request) and it never will.

#### Loaded vs cold

**Loaded** means the model is already in memory and answers instantly. **Cold** means it wakes on your first request — the card shows how long that took when we measured it.

#### Speed

Every model is benchmarked on the real machine before it goes live: tokens per second and time to first token for text, × realtime for transcription, characters per second for speech, and round-trip time for embeddings.

## Integration is a base URL.

Keep your SDK. Point it at the grid. Set your budget and quality preferences in the dashboard — they apply automatically to every request. No special headers or body extensions needed.

curl · chat completion

```
curl https://inpherio.co.uk/v1/chat/completions \
  -H "Authorization: Bearer $INPHERIO_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "gemma4:12b",
    "stream": true,
    "messages": [{"role": "user", "content": "Hello, grid"}]
  }'
```

python · openai sdk, one line changed

```
from openai import OpenAI

client = OpenAI(
    base_url="https://inpherio.co.uk/v1",
    api_key=INPHERIO_API_KEY,
)

stream = client.chat.completions.create(
    model="qwen3:8b",
    messages=[{"role": "user", "content": "Refactor this loop"}],
    stream=True,
)
```

python · text-to-speech (TTS)

```
response = client.audio.speech.create(
    model="speaches-ai/Kokoro-82M-v1.0-ONNX",
    input="Hello from the Inpherio grid.",
)
```

python · speech-to-text (STT)

```
result = client.audio.transcriptions.create(
    model="Systran/faster-whisper-small.en",
    file=open("audio.mp3", "rb"),
)
```

Supported endpoints: `/v1/chat/completions` (text LLMs), `/v1/embeddings` (embedding models), `/v1/audio/speech` (TTS), `/v1/audio/transcriptions` (STT), `/v1/models` (list all), `/v1/messages` (Anthropic format), `/api/chat` (Ollama native). Chat endpoints stream; embeddings and audio return complete responses. Budget and quality preferences are set in your dashboard and applied automatically — no per-request headers needed.

## Stop paying rate-card prices for open models.

Sign up as a developer and start making requests — no waitlist.

[Sign up](https://inpherio.co.uk/login)[Read the spec](https://inpherio.co.uk/how-it-works)
