BroadRouter

About BroadRouter

Updated

BroadRouter is a BroadNet site. It explains the models, prices and setup for BroadNet Inference, BroadNet's OpenAI-compatible API. Accounts, keys and billing live at inference.broadnet.ai. BroadNet has 23+ years in telecom across 17 countries. It hosts its own and open-weight models on owned or leased US hardware, and uses US inference partners for some models and as backup.

What this site is

BroadRouter is BroadNet's website, not an automatic router: it does not route or forward your requests. You choose the model, and your code calls BroadNet Inference directly.

BroadRouter is not an independent review site and not a reseller. It's BroadNet's own guide. BroadNet's own prices are synced from its machine-readable rate card (pricing.json). Other providers' prices are read from each provider's own pricing page, with the source, the date checked and a quote on every row (how we collect them). Every link to sign up goes to inference.broadnet.ai.

Who runs the API

Broadnet Technologies LLC (BroadNet.AI) builds AI products in Arabic and English. BroadNet says it serves the inference itself, with partner capacity as backup, and that it operates its own NVIDIA and AMD GPUs, from edge appliances and workstation RTX cards to B300-class data-center accelerators, and has run about 270 models on them. Its inference privacy statement says its own models and the open-weight models it hosts run on BroadNet-owned or leased hardware in the United States, with partner capacity as backup, and that partner-served models run under zero-data-retention terms with US inference partners. Its Terms of Service name DeepSeek V4.1 Flash as partner-hosted. BroadNet documents the models it rules out as well as the ones it ships: see what it ruled out and why and its GPU hardware and serving approach.

What research has BroadNet published?

BroadNet's research page lists four preprints, each deposited on Zenodo with a DOI. All four are by Mohammad Al Zubaidi, according to their Zenodo records.

These are BroadNet's own studies, published as preprints. They are not peer-reviewed journal papers.

Is there an uptime SLA for the API?

Not for self-serve use. BroadNet's Terms of Service (§7.4) say self-serve "has no availability commitment" and that service levels exist only in a separate written agreement; private endpoints and reserved capacity follow that agreement. The unlimited plans are "delivered best effort under fair use". The 99.99% uptime SLA on BroadNet's site is stated for its Voice AI product, not for the Inference API.

Which regions does BroadNet serve from?

API traffic is processed in the United States by default. BroadNet's inference page says regional processing in MENA or Europe is available on request for customers with data-residency requirements, and private endpoints can be placed in the region you need. BroadNet lists 17 offices, with its headquarters in Dubai; the legal entity, Broadnet Technologies LLC, is in New Jersey.

Editorial standards

This site exists to sell BroadNet Inference, so we hold it to rules that make it useful anyway:

  • BroadNet prices are never typed by hand. Every BroadNet price on the site is inserted at build time from the published price file, and the build fails if a page contains a hard-coded dollar amount. Worked examples are computed from the same file, so they change when prices change.
  • Other providers' prices carry their evidence. 1,671 prices from 41 providers are read from each provider's own pricing page and stored with the source link, the date checked and a verbatim quote. A weekly script re-fetches every source page and flags any price it can no longer find; flagged pages are re-read and the row updated. No router or aggregator feed is used. The full price database is open data.
  • Every outside fact is cited and dated. Competitor prices, limits and policies link to the provider's own page with the date we checked it. Where we could not verify something, we say so instead of guessing.
  • We say when others are cheaper or better. Several articles note providers that charge less per token, or tools that suit a job better. Our honest edge is narrower: flat plans with no token meter, inference content that is not retained, Arabic speech to text and one OpenAI-compatible key.
  • Tools and data show their method. The free tools run in your browser and state their formulas. Token counts use each model's official tokenizer pinned to an exact commit. Our open datasets include their inputs and a script to reproduce them.
  • BroadNet's own claims are attributed. Figures such as the Arabic speech-recognition leaderboard rank are reported by BroadNet and labelled that way.
  • Corrections. If you find an error, tell BroadNet. We fix it, update the page's date and record it in the public fact-check log.

Articles are written by the BroadNet Inference team with AI assistance and checked against the sources listed on each page.

Full details: Editorial policy · Terms of Use · Privacy policy · Refund policy.

Policies

Contact

For private endpoints, reserved capacity or Arabic fine-tunes, see how to reach BroadNet; general questions go to BroadNet's contact page.

Read as MarkdownAsk ChatGPTAsk ClaudeAsk Perplexity