Developers · Vyas language model

A multilingual language model behind an API you already know.

Vyas is the Sonex Labs LLM. It speaks the OpenAI chat completions API, so the client library you use today works unchanged.

Get an API key Read the docs

65,536

Token context window

$0.30

Per 1M input tokens

$0.60

Per 1M output tokens

What you get

Drop-in, multilingual, priced per token.

Works with the OpenAI SDK

Point any OpenAI client library at https://api.sonexlabs.com/v1 with your Sonex API key and use the model vyas. Errors use OpenAI's error shape, so your existing error handling keeps working.

Tools, JSON and streaming

POST /v1/chat/completions accepts OpenAI's request format, including tools, response_format and stream, so agents and structured extraction run without code changes.

Multilingual by design

Vyas is built for the same multilingual world as Pāṇini TTS and Valmiki STT, including Indian languages, so a whole voice pipeline can stay on one platform.

65,536-token context

Prompt and answer together. A longer prompt is refused with context_length_exceeded, and max_tokens is reduced to what fits.

Quickstart

Three lines change, nothing else.

Base URL, API key, model name.

vyas.py

1from openai import OpenAI
2
3client = OpenAI(base_url="https://api.sonexlabs.com/v1", api_key="YOUR_API_KEY")
4
5stream = client.chat.completions.create(
6    model="vyas",
7    messages=[{"role": "user", "content": "मानसून पर दो पंक्तियों की कविता लिखिए।"}],
8    stream=True,
9)
10for chunk in stream:
11    print(chunk.choices[0].delta.content or "", end="")

Building a voice product? Pair Vyas with Valmiki speech to text and Pāṇini text to speech, or let Sonex Labs voice agents run the whole call. Compare every rate on the pricing page.

FAQ

Vyas, answered.

$0.30 per million input tokens and $0.60 per million output tokens. On the monthly plan it is $0.20 and $0.40. Requests refused before the model runs are not billed.

Yes. It implements chat completions, including tools, response_format and streaming. Change the base URL and API key in your OpenAI client and set model to vyas.

Vyas is multilingual, including Indian languages such as Hindi. Test your own language in the playground before going to production.

The same as the rest of the Sonex Labs API: 5 requests per second and 120 per minute per workspace on pay as you go.

Either. Sonex Labs voice agents can run on Vyas, or you can bring your own OpenAI, Anthropic or Gemini key at zero markup.

Send your first Vyas request in a minute.

Create a key in the dashboard, swap the base URL in your OpenAI client, and run it.

Start building free Vyas docs