Text · Vision · Arabic transcription
One API. Your models. Your way to pay.
Build with DeepSeek, GLM, Qwen, Gemma, and Arabic speech-to-text. Pay for what you use, or choose flat-price DeepSeek sessions for your agents.
BroadRouter is BroadNet's website. You pick the model; requests go to BroadNet Inference at https://inference.broadnet.ai/v1.
first_request.py
import os
from openai import OpenAI
client = OpenAI(
base_url="https://inference.broadnet.ai/v1",
api_key=os.environ["BROADNET_API_KEY"],
)
response = client.chat.completions.create(
model="broadnet/gemma-4-26b-a4b",
messages=[{"role": "user", "content": "اكتب رسالة ترحيب قصيرة"}],
)
print(response.choices[0].message.content)request.sh
curl https://inference.broadnet.ai/v1/chat/completions \
-H "Authorization: Bearer $BROADNET_API_KEY" \
-H "Content-Type: application/json" \
-d '{"model":"broadnet/gemma-4-26b-a4b","messages":[{"role":"user","content":"Reply with one short sentence."}]}'| Model | Price | Access |
|---|---|---|
| DeepSeek V4.1 Flash | $0.175 / $0.52 | Top-up |
| GLM 5.3 | $0.78 / $3.46 | Top-up |
| Qwen3.8 27B | $0.175 / $2.17 | Top-up |
| BroadNet Gemma 4 VL 26B | $0.10 / $0.22 | Starting credit |
| BroadNet Transcribe Arabic | $0.0042 / min | Starting credit |
| BroadNet Transcribe | $0.002 / min | Starting credit |
Prices from BroadNet's published rate card, updated 2026-11-14.
Choose the model for your task
Four chat models and two transcription options on one endpoint. Use the exact model ID in your request.
DeepSeek V4.1 Flash
broadnet/deepseek-v4.1-flash- Input
- Text + Image
- Context
- 1,048,576 tokens
- Price per 1M
- $0.175 in · $0.0055 cached · $0.52 out
- Documented
- Reasoning
Top-up requiredSession plans
GLM 5.3
broadnet/glm-5.3- Input
- Text
- Context
- 1,048,576 tokens
- Price per 1M
- $0.78 in · $0.173 cached · $3.46 out
- Documented
- Tools, Structured output, Reasoning
Top-up required
Qwen3.8 27B
broadnet/qwen3.8-27b- Input
- Text + Image
- Context
- 262,144 tokens
- Price per 1M
- $0.175 in · $0.0435 cached · $2.17 out
- Documented
- Tools, Structured output, Reasoning
Top-up required
BroadNet Gemma 4 VL 26B
broadnet/gemma-4-26b-a4b- Input
- Text + Image
- Context
- 32,768 tokens
- Price per 1M
- $0.10 in · $0.22 out
Starting creditHourly key
BroadNet Transcribe Arabic
broadnet/cohere-transcribe-arabic- Languages
- Arabic
- Audio
- WAV or MP3, up to 26,214,400 bytes per file
- Price
- $0.0042 per audio minute
Starting credit
BroadNet Transcribe
broadnet/cohere-transcribe- Languages
- 14, including Arabic
- Audio
- WAV or MP3, up to 26,214,400 bytes per file
- Price
- $0.002 per audio minute
Starting credit
Compare all models · Arabic or multilingual transcription?
Vision means image understanding, not image generation. "Documented" lists features BroadNet's developer reference names for that model (checked 2026-11-15); check the reference for others.
Pay for usage. Or buy a session.
Pay as you go
Every model, billed per token or per audio minute from your account balance. Cached input is billed at the lower cached rate where one is published. Images count as input tokens.
- 30 requests a minute and 4 requests in flight per key
- Starting credit covers Gemma and both transcription models, plus Gemma hourly keys
- DeepSeek, GLM and Qwen need a top-up
DeepSeek V4.1 Flash sessions
Unlimited tokens on DeepSeek V4.1 Flash for the period you buy, at a flat price per session.
| Plan | Price per session | Concurrency |
|---|---|---|
| Full Moon | $60 a month | One request at a time per session |
| Orbit | $15 a week | One request at a time per session |
| Meteor | $1 an hour | One request at a time per session |
Parallel agents need one session each. Plan keys are separate from pay-as-you-go keys, and their requests never draw on your balance. Context limits and the Acceptable Use Policy apply; sessions do not guarantee throughput or uptime. Founders offer: Full Moon at $45 a month per session for the first 3 months with a founders code at checkout, then $60.
Plans and session planner- Request 1
- Request 2
- Request 3
Input tokensOutput tokens
Charge = input tokens × input rate + output tokens × output rate, per 1M tokens
One session runs one request at a time
Agents sending at the same time need one session each
- Agent 1 · Session 1
- Agent 2 · Session 2
- Agent 3 · Session 3
Context limits and the Acceptable Use Policy still apply. A session does not guarantee throughput or uptime.
Pay as you go meters every request by its input and output tokens. A DeepSeek V4.1 Flash session has a flat price for its period and runs one request at a time, so agents that send requests at the same time need one session each.
Not the same as Meteor: a Gemma hourly key is $0.50 for one hour of BroadNet Gemma 4 VL 26B, one stream. Meteor is a DeepSeek V4.1 Flash session at $1 an hour.
Build for Arabic-speaking users

Read Arabic documents and messages with a chat model, and turn Arabic speech into text. The examples below are synthetic.
Extract fields from an Arabic invoice

response = client.chat.completions.create(
model="broadnet/gemma-4-26b-a4b",
messages=[{"role": "user", "content":
"استخرج رقم الفاتورة والتاريخ والمبلغ الإجمالي بصيغة JSON:\n"
"فاتورة رقم INV-1042 بتاريخ 2026-11-02، المجموع 1,250.00 درهم"}],
)Illustrative example
{"invoice_number": "INV-1042", "date": "2026-11-02", "total": "1250.00", "currency": "AED"}Written by us to show the response shape. It is not a recorded model output or an accuracy result. Read invoices from images.
Transcribe an Arabic voice note


- Convert WhatsApp
.oggor.m4ato WAV:ffmpeg -i voice-note.ogg -ar 16000 -ac 1 voice-note.wav - Send it to BroadNet Transcribe Arabic with
language=ar. Files up to 26,214,400 bytes. - Read the transcript from the
textfield.
transcribe.sh
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.wavtranscribe.py
import os
from openai import OpenAI
client = OpenAI(
base_url="https://inference.broadnet.ai/v1",
api_key=os.environ["BROADNET_API_KEY"],
)
with open("recording.wav", "rb") as audio:
result = client.audio.transcriptions.create(
model="broadnet/cohere-transcribe-arabic",
file=audio,
language="ar",
)
print(result.text)Illustrative example
{"text": "مرحباً، لقد وصلت إلى البريد الصوتي، اترك رسالتك من فضلك"}From model choice to first request
Create an account and a key
Sign in at inference.broadnet.ai with Google or an email link. Create an inference key and keep it in an environment variable.
Choose an eligible model
The starting credit covers Gemma and both transcription models, plus Gemma hourly keys. For DeepSeek, GLM and Qwen, top up your balance or buy a session.
Run the code example
Set the base URL and model ID, then send your first request. Quickstarts for Python and cURL.
New accounts start with $5 of credit for Gemma and both transcription models, plus Gemma hourly keys. DeepSeek, GLM and Qwen need a top-up first.

Your content stays out of retained inference records
Inference content is not retained. Necessary account, billing and security metadata is retained as described in our privacy statement.
BroadNet's policy covers prompts, messages, images, audio, answers and transcripts, and says they are not used to train, fine-tune or evaluate models. Some models are BroadNet-hosted and some partner-hosted. If you have data-residency requirements, agree processing locations with BroadNet before use.
Read the full inference privacy statement · What is and is not kept
Questions
Can I try BroadRouter models before paying?
Yes, with limits. New accounts start with $5 of credit for Gemma and both transcription models, plus Gemma hourly keys. DeepSeek, GLM and Qwen need a top-up first. Session plans are bought separately. Accounts are managed on inference.broadnet.ai.
What is a session?
A session is a flat-price allowance of DeepSeek V4.1 Flash for a period: Full Moon, $60 a month; Orbit, $15 a week; Meteor, $1 an hour. Tokens are unlimited for that period, but a session allows one request at a time. Agents working in parallel need one session each. Context limits and the Acceptable Use Policy still apply, and a session does not guarantee throughput or uptime.
Is the Gemma hourly key the same as Meteor?
No. A Gemma hourly key costs $0.50 for one hour of BroadNet Gemma 4 VL 26B with one concurrent stream, and the starting credit can pay for it. Meteor is a DeepSeek V4.1 Flash session for one hour at $1. They are different products with different models.
Are my prompts and audio stored?
Inference content is not retained. Necessary account, billing and security metadata is retained as described in BroadNet's inference privacy statement. Some models are hosted by BroadNet and some by partners; if you need a specific processing location, agree it with BroadNet before use.
Does it work with the OpenAI SDK?
Yes. The API follows the OpenAI chat-completions and audio-transcriptions shapes at https://inference.broadnet.ai/v1. Set the base URL, your BroadNet key and an exact model ID. Then check streaming, tools, structured output, limits and error handling for the model you chose. Migration checklist.
Do plans renew automatically?
Full Moon renews each month on your card. Orbit and Meteor are paid from your account balance; buy another Orbit week before it ends to keep the same key, and a Meteor hour starts at purchase. If you cancel, paid access continues to the end of the period you bought. Taxes and refund rights follow checkout and BroadNet's Terms of Service.
Start building with BroadRouter.
Accounts, keys and billing are managed on inference.broadnet.ai, BroadNet's console. BroadRouter links you there; it does not create accounts itself.
New accounts start with $5 of credit for Gemma and both transcription models, plus Gemma hourly keys. DeepSeek, GLM and Qwen need a top-up first.