Is it actually reachable?

Source

1xAi's own request log: one record per call proxied to an upstream vendor, written at the moment the response is returned.

Vantage point

The gateway's egress inside Iran. These are production requests from Iranian users and applications, not probes fired from a datacenter elsewhere.

Classification

Every request is classified from the upstream's actual response. The exact buckets and the denominator behind the headline rate are declared in the payload itself — schema_version, fields[] and methodology[] — so the definition travels with the numbers instead of living only on this page.

Grain

Per provider, per UTC calendar day, plus rolling 1-hour and 24-hour windows in the current block. A day with no traffic produces no row; absence is not an outage.

Providers

openai, anthropic, gemini, deepseek — the four upstreams the gateway routes to.

Licence

CC-BY-4.0. Attribution line: “Source: 1xAi provider reachability dataset — https://1xai.ir/status”.

The dataset URL is stable, unauthenticated and CORS-open, and the JSON carries its own methodology, field descriptions and caveats, so a reader who finds the file without the page still gets the qualifications with the numbers.

# the whole document, no key, CORS open
curl https://1xai.ir/status/dataset

# just the daily series, 90 days, as CSV
curl "https://1xai.ir/status/dataset?days=90&format=csv" -o reachability.csv

The CSV is generated from the same historyarray the JSON exposes, with the same rows, order and rounding, so diffing the two formats finds no discrepancy. Its header row names the columns and the JSON's fields[] defines each one, which is the pair to read before quoting anything.

import httpx

doc = httpx.get("https://1xai.ir/status/dataset", timeout=30).json()

print(doc["generated_at"], doc["vantage_point"], doc["license"])

for row in doc["current"]:                      # rolling 1h / 24h windows
    print(row["provider"], row["state"], row["ok_pct_24h"], row["reqs_24h"])

for row in doc["history"]:                      # daily series
    if not row["partial"]:
        print(row["day"], row["provider"], row["ok"], row["requests"])

• —It measures reachability through 1xAi, not a vendor's global health. A depressed rate can mean our routing, our upstream account, our rate limit or our egress path.
• —Volume differs enormously between providers. A provider with a few dozen requests in a day has a noisy percentage, so read the requests column before quoting the rate.
• —Request volume reflects what our customers chose to call. It is a usage signal for this gateway, not market share for a vendor.
• —It is first-party operational data. Not an independent survey, and not audited by a third party.
• —Historical rows are recomputed from the log on every request, so a figure stays reproducible. But the current day's row keeps moving until the day ends, and a revision of the schema can change how a rate is derived. Quote the schema_version alongside the number.

Any gateway can print "99.9% uptime" on a landing page, and none of them can be checked. The alternative is to publish the counts instead of the claim: requests and successes, per provider, per day, with the definition of success stated and the failure modes listed. If you are deciding whether to put a production dependency behind an intermediary, that is the artefact worth reading, and it is the same artefact whether you are a customer, a journalist or a researcher.

Reachability from Iran is also not a constant. It is the thing this whole product exists to manage, so it is measured continuously and shown in public, including on the days it looks bad.

1xAi is an independent gateway. It is not affiliated with, endorsed by, or a reseller for OpenAI, Anthropic, Google or DeepSeek. Requests are relayed to those providers' official APIs, so each provider's own usage policies apply to what their models will and will not do, and their model availability is theirs to change. Current reachability is published on the status page.