Thinking Machines Lab Outage History

21 incidents recorded over the last 365 days, 21 of them resolved. Durations are measured from when PulsAPI first saw the problem to when it cleared, which is usually longer than the vendor's own figure.

Live Thinking Machines Lab status100.00% uptime history

Every Recorded Thinking Machines Lab Incident

Thinking Machines Lab incidents, newest first, with date, duration and status.
DateIncidentDurationStatus
Oct 28, 2026Degraded sampling for DeepSeek-V4.1-FlashBetween 10:44 and 12:47 UTC, sampling requests for DeepSeek-V4.1-Flash stalled and did not return results. Other models and the OpenAI-compatible endpoint were not affected. A fix is deployed and capacity is recovering…-2 minResolved
Oct 23, 2026Tinker sampling downThe issue has been fully resolved.7 minResolved
Oct 10, 2026Tinker backend calls failing with 5xx errorsThe issue has been fully resolved.2h 39mResolved
Oct 8, 2026Billing backend outageThe incident is fully resolved.1 minResolved
Oct 2, 2026sampling broken for gpt-oss-20b and gpt-oss-120bThe incident is fully resolved.47 minResolved
Sep 7, 2026LoRA rank <8 may fail training on MoE modelsThe issue has been fully resolved.17h 14mResolved
Aug 11, 2026Temporary outage due to billiing backendThe issue has been fully resolved.6h 33mResolved
Aug 9, 2026All sampling was down 6:36PM PT - 7:01PM PT on 08/08The issue was resolved.61h 53mResolved
Aug 6, 2026Training on Inkling could fail with variable-shape loss inputs — SDK 0.24.1 requiredSDK version 0.25.0+ is required to use Tinker API.753h 4mResolved
Aug 5, 2026Getting model info can fail for old modelsThis is fully resolved now.48 minResolved
Aug 5, 2026Tinker sampling slow or stuck for most modelsThe issue is fully resolved now.27 minResolved
Jul 28, 2026Tinker sampling failuresThe problematic code was reverted, the issue is fully mitigated.54 minResolved
Jul 25, 2026Degraded Nemotron and Qwen3.5/6 SamplingWe are rolling out a fix that will restore any lingering reduction in prefill throughput11h 9mResolved
Jul 24, 2026Degraded Training CapacityWe have deployed a fix and capacity has recovered6h 34mResolved
Jul 22, 2026Tinker SDK may be getting stuck on MacOSSDK version 0.25.0+ is required to use Tinker API.1130h 30mResolved
Jul 21, 2026Training DegradedThis issue has been resolved54 minResolved
Jul 16, 2026Tinker fine tuning sampling slowThe issue is fully resolved now.1h 5mResolved
Jun 30, 2026Several Models DownAll models should be back up now!10 minResolved
Jun 11, 2026Delayed TrainingAll training should be running normally50 minResolved
May 14, 2026API unstableAPI has recovered. Root cause is identified and we will push that fix out soon.2h 18mResolved
Apr 21, 2026Degradation in checkpoint deletionWe've pushed a fix for this issue.193h 50mResolved

Hear about the next one while it is open

Everything above is a record written after the fact. We can email you when the next Thinking Machines Lab incident opens instead, at the point where knowing still changes what you do.

Free, no password, no card. Unsubscribe in one click.

How to Read This Thinking Machines Lab Incident Log

Each row is an incident PulsAPI observed, not a summary written afterwards. The duration is wall-clock time between the first failing check and the first clean one, so it includes the window before Thinking Machines Lab acknowledged anything. Vendor post-mortems typically measure from acknowledgement, which is why their numbers are usually shorter.

Incidents still open have no duration yet and are listed as ongoing rather than being given a running total. The archive covers the last 365 days; anything older has aged out of the window rather than never having happened.

The longest single Thinking Machines Lab outage in this window ran 47d 2h, against a mean recovery of 4d 13h. If you depend on Thinking Machines Lab in a customer-facing path, the longest figure is the one to design around. The mean is what happens on a normal bad day; the maximum is what happens on the worst one.