Public infrastructure observatoryRecord mistral
Mistral AI
Mistral AI is observed by probing its public https://status.mistral.ai from 1 observation region. RELIASTRA measures the HTTP behaviour of the listed endpoint - its state is not a statement about every service Mistral AI operates.
The most recent observation did not receive the expected response from this endpoint.
Observations have been arriving about every 5.0m in this window. The bar advances against that interval.
Elapsed time since the last observation is being measured.Is Mistral AI down?
The endpoint Mistral AI is observed at is not responding: the latest RELIASTRA check of status.mistral.ai did not receive the expected response. Whether Mistral AI’s services are affected is a separate question this record does not answer alone.
- Latest observation: 26 Sep 2026 20:53:28 UTC - HTTP 403 in 84 ms from us-east.
- Last 24 hours: 44 observations, availability 0.00%, mean response 84 ms. Checks arrive about every 300 seconds per region.
This record measures the HTTP behaviour of status.mistral.ai - the endpoint RELIASTRA probes for Mistral AI. It is not a statement that every Mistral AI service is operational, and an operational endpoint is not a statement about your own integration, your traffic path, or your API key.
Vendor-reported status is separate: RELIASTRA does not ingest, mirror or reconcile Mistral AI’s own status reporting, so the two records can legitimately disagree. Read both; only one of them is independent of Mistral AI.
AI infrastructure hub · shareable direct answer to this question · measurement methodology
- Category
- ai
- Observed endpoint
- status.mistral.ai
- Observation regions
- us-east
- Measured by
- RELIASTRA probes
Summary of this record
Mistral AI is observed by RELIASTRA as an ai dependency under continuous observation. Requests are issued to https://status.mistral.ai from us-east, about every 300 seconds per region. Each observation records the response latency, the HTTP status code or transport error, and a UTC timestamp. The last observation recorded was 26 Sep 2026 20:53:28 UTC.
Over the last 24 hours RELIASTRA recorded 44 observations, an availability of 0.00% and a mean response of 84 ms; over 30 days, 0.00%. RELIASTRA lists 1 public incident record for this dependency. Availability, latency and incident history on this page are measured by RELIASTRA probes, not reported by the vendor, and describe the observed endpoints rather than every service Mistral AI operates.
Current observation
- Latency
- 84ms
- Status
- 403
- Observed (UTC)
- 26 Sep 20:53:28
These observations come from a single observation point, so they can show that the endpoint answered or did not - they cannot corroborate a vendor-wide outage. A region reporting no response is a fact about this path, at this minute.
Observed state and availability
- Observations
- 11
- Mean latency
- 71ms
- 95th percentile
- 120ms
- Observations
- 44
- Mean latency
- 84ms
- 95th percentile
- 182ms
- Observations
- 44
- Mean latency
- 84ms
- 95th percentile
- 182ms
- Observations
- 44
- Mean latency
- 84ms
- 95th percentile
- 182ms
- Observations
- 44
- Mean latency
- 84ms
- 95th percentile
- 182ms
- Observations
- 44
- Mean latency
- 84ms
- 95th percentile
- 182ms
Every window is computed by the measurement API over stored observations at read time. Latency figures are milliseconds; a mean of zero means no successful response was recorded in the window and is printed as no data.
Historical telemetry
Initialising observation series…
Requesting aggregated observations from the measurement API. No values are shown until they arrive.
Viewing 90 days from us-east. Other ranges: 24 hours, 7 days, 30 days.
Observation network
- Observed cadence
- 300s
Cadence is estimated from the spacing of observations in the last hour, not read from a configured schedule: the public API does not expose the check interval, so this record reports the interval it can verify. An interval longer than the chart resolution is read from the gap between observations, not from their density.
Observed incidents
- Window (UTC)
- 26 Sep 2026 17:10:56 → open
- Duration
- ongoing
- Status
- Open
- Evidence
- not published
Times are UTC. An open incident has no resolution time because RELIASTRA has not yet observed a qualifying recovery. Every row links to a published record at its own address.
Evidence records
No published evidence for this dependency
Evidence records appear here when an organization monitoring mistral on RELIASTRA publishes the report for an incident. Until then the incident history above is the public record, and the underlying observations remain in the accounts that collected them.
How RELIASTRA knows this
- Source
- Scheduled workers issue real HTTP requests to the endpoints listed at the bottom of this record, from us-east. Where a listed endpoint is Mistral AI's public status site, the probe measures that site's own availability and response time as an HTTP service; the status text it publishes is not read, parsed, ingested or reconciled.
- Interval
- About every 300 seconds, measured from the spacing of observations in the last hour. The configured interval is not exposed publicly.
- Success
- The probe records a success when the endpoint returns the response it is configured to expect - HTTP 200 for every public record, within a 15-second deadline, following at most five redirect hops. A slow but correct response is latency, not downtime.
- Failure
- A timeout, a transport error, a redirect the security policy blocks, or any status code other than the expected one. Each is stored with its error type so the causes stay distinguishable.
- Incident opens
- For public records like this one, nothing: the public pipeline stores observations and does not open incidents. Customer monitoring under the deployed single-origin topology opens an incident after two consecutive failed checks from one observation point; with a genuine multi-origin fleet, two or more independent points must agree inside a 60-second window. Two labels from one worker never count as agreement.
- Incident closes
- Two consecutive successful observations from the same observation point - or, under a multi-origin fleet, from at least two points - so flapping does not open and close a record repeatedly.
- Availability
- Successful observations divided by all observations in the window. A window with no observations is insufficient data, never 100%.
- Latency
- Mean and 95th percentile of response times in the window, in milliseconds, across every observation in it. The chart breaks the line where a bucket has no successful response.
- Freshness
- Rendered on the server and revalidated every 60 seconds. An observation older than 15 minutes marks the record stale rather than healthy or down. All timestamps are UTC.
- Not vendor status
- Mistral AI's own status reporting is not ingested, mirrored or reconciled. The two can disagree - a measured endpoint can be up while services are degraded, and vice versa. Read both.
- Limits
- These figures describe the listed endpoints, observed from the listed region labels. They are not a statement about every service the vendor operates, about specific API routes or models, or about your integration.
This is not a status page
- Endpoint vs service
- The endpoint observed here (status.mistral.ai) is Mistral AI's public status site. This record measures that endpoint's HTTP behaviour - whether it answers, how fast, with what status. A status site can be perfectly reachable while the services it reports on are degraded, and unreachable while they are healthy. This record answers neither question; it answers the first one exactly.
- Official vendor status
- This is not Mistral AI's official vendor status page, and it is not affiliated with or endorsed by them. It is an independent measurement of the endpoints listed in this record.
- Vendor-reported state
- RELIASTRA does not ingest, mirror or reconcile vendor-reported state. Nothing here is copied from a status page, an aggregator or a vendor API. Every figure is derived from observations RELIASTRA issued itself.
- Why that matters
- A status page is written by the party whose reliability it describes, and is updated when that party decides to update it. An independent record is written by the observer. Where the two disagree, both are worth reading - but only one of them is yours to keep.
- Scope
- These observations describe the listed endpoints from the listed region labels. They are not a statement about every service Mistral AI operates, or about your specific integration.
Dependency information
https://status.mistral.aiHTTPS · host status.mistral.ai · path /- Regions
- us-east
- Active
- yes
- Endpoint state
- down
- Record identifier
- mistral
- Category
- ai
- Record opened
- 26 Sep 2026
- Visibility
- public
Your own record
Mistral AI is one dependency. Your product has a list.
Observe the services your product calls. Attribute incidents. Keep the record.