BroadNet
BroadRouter by BroadNet · New

One API for DeepSeek, GLM, Qwen, Gemma and Arabic speech

BroadRouter is BroadNet's developer hub for BroadNet Inference, an OpenAI-compatible API for four chat models and two speech-to-text models. Keep your SDK, change the base URL, and pay per token or a flat price per session.

New accounts start with $5 of credit, no card needed, for Gemma and both transcription models. Try it without an account (opens in a new tab)

Example
import os
from openai import OpenAI
client = OpenAI(
base_url="https://inference.broadnet.ai/v1",
api_key=os.environ["BROADNET_API_KEY"],
)
reply = client.chat.completions.create(
model="broadnet/deepseek-v4.1-flash",
messages=[{"role": "user", "content": "Summarise this ticket in one line."}],
max_tokens=1500,
)
print(reply.choices[0].message.content)

Your existing OpenAI SDK code, pointed at BroadNet Inference: change the base URL, key and model ID.

models on one OpenAI-compatible endpoint
6models on one OpenAI-compatible endpoint
tokens of context on DeepSeek and GLM
1Mtokens of context on DeepSeek and GLM
an hour for unlimited DeepSeek V4.1 Flash
$1an hour for unlimited DeepSeek V4.1 Flash
prompts, files or answers retained
0prompts, files or answers retained
What BroadRouter is

A developer hub, and the API behind it

BroadRouter.com explains the models, prices and setup. Your code calls BroadNet Inference directly at inference.broadnet.ai/v1, where accounts, keys and billing live. It is not a reseller and not an automatic router: you choose the model, and BroadNet serves it.

Models and synced prices

Six models with exact IDs, context windows and prices synced from BroadNet's published rate card, plus side-by-side model comparisons.

Compare the models (opens in a new tab)

Setup guides for 30+ tools

Step-by-step setup for the OpenAI SDKs, LangChain, LlamaIndex, n8n, coding agents and chat apps, with the exact fields to fill in and the errors to expect.

Browse the integrations (opens in a new tab)

Free tools and open data

Browser tools such as an LLM cost calculator, token counters and an Arabic token counter, plus open datasets on LLM prices and tokenizer efficiency under CC BY 4.0.

Open the free tools (opens in a new tab)
Models

Four chat models and two speech models

Send the exact model ID in your request. Prices are in USD from BroadNet's published rate card of 14 November 2026; the live rate card is the price of record.

  • DeepSeek V4.1 Flash

    broadnet/deepseek-v4.1-flash
    Top-up
    Input
    Text and images
    Context (tokens)
    1,048,576
    Price (USD)
    $0.175 / $0.52 per 1M tokens, input / output$0.0055 cached input

    Best for: Coding agents, long documents and anything you want to run unlimited on a flat plan.

  • GLM 5.3

    broadnet/glm-5.3
    Top-up
    Input
    Text
    Context (tokens)
    1,048,576
    Price (USD)
    $0.78 / $3.46 per 1M tokens, input / output$0.173 cached input

    Best for: Tool calling and structured output over very long context.

  • Qwen3.8 27B

    broadnet/qwen3.8-27b
    Top-up
    Input
    Text and images
    Context (tokens)
    262,144
    Price (USD)
    $0.175 / $2.17 per 1M tokens, input / output$0.0435 cached input

    Best for: Arabic text and images, including Gulf and Egyptian dialects.

  • BroadNet Gemma 4 VL 26B

    broadnet/gemma-4-26b-a4b
    Starting credit
    Input
    Text and images
    Context (tokens)
    32,768
    Price (USD)
    $0.10 / $0.22 per 1M tokens, input / output

    Best for: The lowest price per token, trying the API on the starting credit, and flat-price hourly keys.

  • BroadNet Transcribe Arabic

    broadnet/cohere-transcribe-arabic
    Starting credit
    Input
    Arabic audio
    Price (USD)
    $0.0042 per audio minute

    Best for: Arabic audio: voice notes, calls and dialects.

  • BroadNet Transcribe

    broadnet/cohere-transcribe
    Starting credit
    Input
    Audio in 14 languages
    Price (USD)
    $0.002 per audio minute

    Best for: Transcribing speech in any of 14 languages at the lowest price per minute.

Vision means reading images, not generating them. DeepSeek, GLM and Qwen are reasoning models: reasoning counts toward max_tokens, so set it to 1,000 or more. Live rate card (JSON) (opens in a new tab)

Pricing

Pay per token, or a flat price per session

Every model is billed per token or per audio minute from your balance. For agents that run all day, DeepSeek V4.1 Flash is also sold as unlimited sessions.

Pay as you go

Every model, metered per token or per audio minute from a prepaid balance. Cached input is billed at the lower cached rate, and images count as input tokens.

  • 30 requests a minute and 4 in flight per key
  • Up to 10 live keys per account, each valid for one year
  • Gemma hourly key: $0.50 for one hour, one stream
See per-token pricing (opens in a new tab)

Unlimited DeepSeek V4.1 Flash

No token meter, no daily cap and no weekly limit. A session runs one request at a time, so agents working in parallel need one session each.

  • Most chosen for agents
    Full Moon

    $60 a month

    per session

    Renews monthly. Cancel anytime.

  • Orbit

    $15 a week

    per session

    A one-week project, paid from your balance.

  • Meteor

    $1 an hour

    per session

    Try it on a real task. Starts at purchase.

Founders offer: Full Moon at $45 a month per session for the first 3 months, with a founders code at checkout.

Plan keys are separate from pay-as-you-go keys and never draw on your balance. Self-serve use is best effort under fair use, with no uptime commitment; service levels come with a written agreement.

When does a flat plan pay off?

At the typical 100 tokens a second, one session kept busy around the clock can generate about 259 million output tokens in 30 days, which would cost about $135 at the per-token output price alone. An agent that runs a few hours a day often costs less per token, so check your own usage first.

Built for Arabic

Arabic speech, documents and chat

BroadNet has served Arabic-speaking markets since 2003, and this API puts Arabic first: dialect speech recognition, Arabic documents and models chosen for Arabic text.

Arabic transcription: models, formats and cost (opens in a new tab)
  • Every major dialect

    BroadNet Transcribe Arabic handles Gulf, Egyptian, Levantine, Maghrebi and Modern Standard Arabic, returns Arabic script, and costs $0.0042 per audio minute, about $0.25 an hour.

  • On a public leaderboard

    The Open Universal Arabic ASR Leaderboard listed BroadNet's Arabic model #4 of 39 systems on 15 November 2026. BroadNet reported it #2 of 37 in September 2026.

  • Voice notes in one API request

    Convert a WhatsApp .ogg voice note to WAV with ffmpeg and send it with language=ar. Files of up to 26,214,400 bytes are accepted.

  • Invoices and receipts to JSON

    Send a photo of an Arabic invoice or receipt to a vision model and get the number, date and total back as JSON.

  • Fewer tokens for the same Arabic

    The same Arabic text needs 1.31× the English token count on Qwen3.8 27B and 1.62× on DeepSeek V4.1 Flash, by BroadRouter's open tokenizer study. Pick the model with your language in mind.

Example
ffmpeg -i voice-note.ogg -ar 16000 -ac 1 voice-note.wav
curl https://inference.broadnet.ai/v1/audio/transcriptions \
-H "Authorization: Bearer $BROADNET_API_KEY" \
-F model=broadnet/cohere-transcribe-arabic \
-F language=ar \
-F file=@voice-note.wav
Privacy

Your prompts and files are not kept

BroadNet describes zero data retention as built into how requests are served, not only stated in policy.

  • Zero data retention

    Prompts, images, audio, answers and transcripts are processed for the request, then not retained or logged. They are never used to train, fine-tune or evaluate models, and BroadNet checks its gateway configuration every hour.

  • What is kept

    Only content-free records needed to run the service: request IDs, timings, token counts and cost, plus account, billing and security data.

  • Where it runs

    BroadNet's own and hosted open-weight models run on BroadNet-owned or leased hardware in the United States. Partner-served models, such as DeepSeek V4.1 Flash, run with US inference partners under zero-data-retention terms. Regional processing in MENA or Europe is available on request.

  • Keys you control

    Inference keys call model routes only and are shown once. Lock them to your servers with IP allow-lists, and revoke a leaked key at once.

What is kept and what is not (opens in a new tab)
Quickstart

From sign-up to first request

  1. 1

    Create an account and a key

    Sign in at inference.broadnet.ai with Google or an email link, then create an inference key and keep it in an environment variable.

  2. 2

    Choose a model

    The starting credit covers Gemma and both transcription models. For DeepSeek, GLM and Qwen, top up your balance or buy a session.

  3. 3

    Change the base URL

    Point your OpenAI SDK at https://inference.broadnet.ai/v1, set the model ID and send your first request.

  4. https://inference.broadnet.ai/v1 · BroadNet Inference serves the chat completions and audio transcriptions endpoints, not the Responses API. Send the key only in the Authorization: Bearer header.
Example
import os
from openai import OpenAI
client = OpenAI(
base_url="https://inference.broadnet.ai/v1",
api_key=os.environ["BROADNET_API_KEY"],
)
reply = client.chat.completions.create(
model="broadnet/deepseek-v4.1-flash",
messages=[{"role": "user", "content": "Summarise this ticket in one line."}],
max_tokens=1500,
)
print(reply.choices[0].message.content)

Questions about BroadRouter

What is BroadRouter?

BroadRouter (broadrouter.com) is BroadNet's developer hub for BroadNet Inference, an OpenAI-compatible API for large language models and speech-to-text. It publishes the models, prices, setup guides, free tools and open data, and your code calls BroadNet Inference directly at https://inference.broadnet.ai/v1.

Is BroadRouter an LLM router or aggregator?

No. Despite the name, it does not route or forward requests between providers. You choose a model ID and call BroadNet Inference, a single provider, directly.

Which models can I use?

Four chat models (DeepSeek V4.1 Flash, GLM 5.3, Qwen3.8 27B and BroadNet Gemma 4 VL 26B) and two speech-to-text models (BroadNet Transcribe Arabic and BroadNet Transcribe, which covers 14 languages). DeepSeek, Qwen and Gemma also read images.

How much does it cost?

Pay as you go starts at $0.10 per million input tokens on Gemma and $0.002 per audio minute for transcription, on the rate card of 14 November 2026. Unlimited DeepSeek V4.1 Flash costs $60 a month, $15 a week or $1 an hour per session.

Can I try it for free?

Yes. New accounts start with $5 of credit, with no card needed, for Gemma, both transcription models and Gemma hourly keys. DeepSeek, GLM and Qwen need a top-up first. You can also try Arabic transcription and Gemma chat in BroadNet's browser playground with no account.

Does BroadNet keep or train on my data?

No. Prompts, images, audio, answers and transcripts are not retained or logged, and are never used to train, fine-tune or evaluate models. BroadNet keeps only content-free records such as timings, token counts and cost.

Will my existing OpenAI code work?

Yes, if it uses chat completions or audio transcriptions. Change the base URL to https://inference.broadnet.ai/v1, use a BroadNet key and a BroadNet model ID. The Responses API is not served.

How do unlimited sessions work with several agents?

A session runs one request at a time with no token meter. Count the agents that call the model at the same moment, including sub-agents, and buy that many sessions. Plan keys work only with DeepSeek V4.1 Flash.

Is there an uptime SLA?

Not for self-serve use. Private endpoints, reserved capacity and Arabic fine-tunes are available with a written agreement that sets the service levels. Contact BroadNet to discuss one.

Start building

Send your first request in five minutes

Create a key, change one line and run DeepSeek, GLM, Qwen, Gemma or Arabic transcription. For larger workloads, BroadNet offers private endpoints, reserved capacity and Arabic fine-tunes.

Talk to BroadNet about private endpoints